FOR OFFICIAL USE ONLY

Transcription

1 Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Report No AFR Analysis Plans for Understanding the Social Dimensions of Adjustment July 1990 SDA Unit Africa Region FOR OFFICIAL USE ONLY Dot~imtof the Woddbok 1 % IN~~~ ~This docun-ent la restricted-distribvtimn -and may be used by recipients i only n=.e,r' te _anceof their offikial ffuties.4tsoe n not Qtmwse, b scfo Wo Wo Bank - auth,rizton. be di l ; r ; v

2 W~R Q L USB t ULY Analysis Plans for Understanding the Social Dimensions of Adjustment Tne World BN Wasio, D.C. llb 4_18 ma temd d amnd yrbe useyd b soenl j In 6epece Mof thefir 1Iw hdhdebedbe&&iwad wmw&b" mltdb

3 This Vohume on SDA Analysis Plans has been edited by Mrco Perroni, under the overall guidance of Lionel Demery, Chrsfta Giotaert and Michel NoeL Authordip of the Chapters contaned in this Volume is as folows: Chapter 1 (Analysis Strategy): Marc Ferron and Chrstiaan Grootaert Chapters 2 and 3 Macro-Meso Linkages and the Poverty Profile): Lioe Denme, Chapter 4 (Employnt and Eaing): Tony Addison; Chapter5 Health). Mark Pitt; Chapter 6 (Education): Jere Be an; Chapter 7 (Fod Security arnd Nutritio;): John Hoddint Roger Hay and Graham Eele; Chapter 8 (Women and Strucural Adjustment): Paul CoUier, Capter 9 oldersx subidtted by IFAD under the authos*dp of Joseph Mulen and Riard Pearce (ttds Chapter was revised by Leuma Merid); Chapter 10 (The Priority Survey Analysis Plan) Lionel Demery and Christiaan Grooter The Techmca Appedix on imcome and expenditure awegates was written by Andrew McKay and Jeffemy Round. Jorge Wong-Valle assisted in preparing the Volume. The draft SDA Analysis Plans have benefitted from conments received during an Inter-Agency Technical Workshop on the SDA Infmation System which was held at the World Bank in November The Analysis Plans were further reviewed during an authors' workshop held at Woodstock (UK) in Febry lhe Volume also benefited from a,nrn-nts raceived fromthe mebersof a World Bankr eview anelchaired by Steve OBrien, and composed of Ridhard Westebbe, Nancy GMespie, Ravi Kanbur, Martha Ainswort, Peter MooCk, CoDelis Tuinnurg Subbarao K}aidhi, Osita Ogbu, and Frederic touat.

4 Table of contents UxUive summa I Overew 1 Analyzirg tme me-eonomc effects of structurl adjustment 2 The pove profile 3 tnbmenandnmnieffects 3 Analyzing human resource effects. Health 3 Anlyzing hunan resoure effects Education 4 Househld food security and nutritlon 4 The Impact of adustment on women 5 Saliholdersand struc l adiustment 5 An daiysis pan for the SDA piority svey 6 Introduction and ovaview 7 chapter On An analysis stratgy far the soial diensions of adstmet 9 Adjustme and distribution: Cownderatons towd an ualytial approach 9 Identfyit the key socidal dimeons 11 Data soures 12 An anals strate for the sl dienons of adjustent 15 Cbapter Two: Analyzing the meso-economic effecs of stuca adjustment 23 InIrvouctio 23 The context of mnaco-meso analysis 24 A review of the neso effects of adustment 26 Essenatial betkrund Information 27 ExportIe imrts and nontradable 29 The processof adjust*me 30 Expendare4wit6 effects 32 Expenditue-redudngr ffects 36 Economy-wide moddu techqes 38 Concludingobsevations 42' Notes 42

5 a. Jar Sr Sr 118Sr Sr gw 3' IT; II rii'qjri fgj i[pr dl 4, - -p. ii r'"ii I J r.iij Ii 2; I i I I C j 4, 0 0 A. *ffii jttii I I to ii - iii I ; I

6 Chapter Nin Smabodeas and stuctral adjustnmt 165 ntoduction 165 Macro-meso lnkags 166 Micromeso analysis 169 Polcy perspective 180 Notes 181 ChapterTer Ananalysis plan for the SDA prioity svey 1 Introduction 183 Data content 184 Target groups and sodoeconomic groups 185 ask nds 188 Employment 194 Income and assets 200 Conclusion 206 Technial Appendix Coneps and esdmation of household haoma and expenditures within a sytem of household accoun 207 Intrduction 207 Household ammnts: Conceptual basis 208 The estination of household accounts from the SDA inteh_ted survey 215 Condusions 225 An Detled methodolog for acent acot agpgr s Bibliography Tables Table 1.1: A hierbrhal Information system for the SDA Table 2.1: Meso-level vaiables utiized in househd data analyss Figure 2.1: Macro-iwso-mlcro anlwys FIgu 7.1: LInks between macro policy and food secluity re8.1: -Pionof activitiesin theraw. -' r.n of Figur 8±2: Income and pimaty sdhool non-enrollm dfmut Boe Box 2.1: An example of reltve-pice caculations Box 3.1: Assessing the groupns Box 3± In3 oe sours: an lustration Box 5.1: The health production (unction Box 52: Ca d utilitym izaon Box 53: The reduced-form demand equation for health Box 5k Techni iissues in regresion analyis Box 7.1: Css tabulation of total food expenditure by expediu Box 72: Expenditueelscides of foods in Ghan yg p Box 73: EstUrng calo*4rincomelasticities Box 7: Estimating price esticities Box 7.5 Example of the analyas of nutritional data Box 7.6: t the dete d inants of antaopom stats In Gha Box &1: Using swvey data to re-agregate Box 8 Presntp genderdffeens in illness Box 83 Gender diffen in the allocation of labor dne

7 ~~~~. ~~~~~~~~a w ~ ~ ~ ~ ~ ~ sr~ ~~o r I-~ ~~~~~~~~~~~~~~~~~~~~~~~~A.. i~~~~~~~~~~~~~~~~~~~~~~~~~~~~~i

8 FhneA.12: Easof household non-farm enterprises for poor and non-poor households Frame 4.13e lte structue of wage diffrentials by educaton level and age Frme 414: Partcipation in wage-employmnt independent varlables detmining (I) the decision to participate or not (II) the number of hours supplied by participarns Frame 4.15: Wependent vafiables for an eanings function Frame 4.16 The determinants of the profits of household non-farm enteprises Frame 5.1: Weight-for-height of children by age and program availability Frame 6.1: Suggestive list of meso variables for SDA analysis plan for education Frane 6.2: SumnmyofschoohngcbaerW csfromsdasuve.range,ndianmean, stundard deviation, and number of observations for following variables by ageand sex Frame 6.3: Regression variables for deteminants of etucation Frame 8.1: Relating macroeconomic secomrs to activities idented in surveys Frame 8.2: Wage rates and the distribution of the laborfwe by gerder Frame 8.3: The allocation of land by sector and gende Frame 101: The distribution of the population by age, sex, and SEC Frame 10.2: Distribution of households across SEGs Fame 103: School enrollment rates by level of schooling sex and SEC Frame 10.4: Educational drop-out rates by socio-economic group Fran* 10.5: Percentage of children not curently attending school for financa and logistic reasons, by level of schooling and soclo-onomic goup Fame 10.6: Age/grade mnsmatches as a proportion of total enrollments, by SEC, education levels and sex Fme 10.7: Age-rade matrix for SEG 1 Frame 10.8 Mean houshold education expenditure per crently enrolled child, by socio- economic goup and region Frane 10.9: Incodence of health consultations by SEG, age and sex Framne 1010: Distribution of health consultations by type of consultation and socioeodnomic group Frame 10.11: Mean percapita household expenditureon healthbysoca-economic group and region Frame 10.12: Average health cost per consultation by type of consultation and socioecono group Frme 10.13: Penage of children (aged 3-60 moniths) indicating dignfcantly low weight for height, by SEG and region Frame 10.14: Distribution of households by house tenure and socio-economic group Frame 10.15: Distributionofchangesintenurial a tsforhouseholddweingby socio- economic gup Frame 10. Distribution of households by source of water and soclo-econondc group Frame 10.17: Disroution of chage in water sources by so-economic group Frame 10.18: Distbution of households by type of cooking fuel and soco-economic goup Frame 10.19: Disibuftion of chanlges in cooking fuel by soci-economic group Frame 10. Dstribution of individuals (age7and older) by min economic activity and SEG Frame 1021: Unemployment rate by age/sex and SEC Frae 1022a: Main occupation of the spouse in rural SECs Frame 10.2Zb Main occupation of the head of household and spouse in urban SEGs Frame 10.23: Pattems of occupational changes in e last x years, by occupation Frame 1024: IWcdence of occupational changes in the last x yeaws, by curent main occupation and SEC ix

9 Frame 1025: Wdence of secondary job holdin, by SEG and current main occupation Frame 126: Patten of secondary,ob holding, by curt main occupatio Frame t027: Mean shes of income source by socdo4conom grup Frame 10.2& Pwportion of househlds e chnges in income by incoe souuve and so_o.economicgroup Frame 10.29: Mean Income-change scoes by sodoeonomndc group Frame 1030: Mean wage income of household head and spouse, by SEG and sector Frame 1031: Proporton of fannrsesxperiencngchanges in production and area planted bycrop Frame 1032: Reasons for change in ea planted by direction of chnge and crop Frame 10.33: Proportion of faring houslds ing inputs by type of Input and main crop Frame 1034: Number of non-farm entprises by sector and SEG Frame 1035: Proportion of enteprises eo ncin output and employmet *mws by sector Frame 1036: Asset owneship Frame A.l: Aggregates of the houseold cumrnt account Fam A2: The agriultural producion account in sdematic fonn Frame A.3: The production account of a non-farm enterprise Frame AA *:hemutic rprsetation of the househod capital account x

10 Executive summary Overview distributional and welfare asquee of sabi- -iato and adtd teimely develop The SDA Pgr, launched jntlryb the mentof opea e e ns n priority Aft:anDevdopmntBank,UNDPandeWorId policy areas) is compex In a nuthl, * SlDA -k, wit te World Bank as Exeut Agency, resarch task to trl acro-co relation and funded by a number of mudtilateral and bi- ships in order to understand the processes laterl donors, supports partcipat Afican through,whichthe efferts of mroemic polgodvmnts(30 to date) In their efft to assess icylefbn (Orext cn e and monitor e secwal and grco- households This less sa rward sequenes of mareconomi adjutmet Te becase of teretical lmitatons and dat concontibuions of the SDA P dct incude activi- straints at the hoold level in Sub9aha dte ward improved, povefty-cosus mcro- Aficia. Macr- and micrae onon*s lack wheoeconomic and sectralc licy (pol- retcal intqrafrio, and macro-micro analsis s icy sudies and te izan of fora to fail- complated by he fact at )() private behavior tate exchanw betw poli analysts and deci- may offset to Itded effects of adjstment slon-makrs); the of natoa sta- poliy on pricds and th avai ty of goods; (Ui) stca systems with a view to Fra the netmaketftsuydepend on ouer!acti houwhold atd conumuiy4evel infmation e response with hory offerg Ulttle guiuance re quired by te abuve*hedesignadmpile grdhg thdr rlative mn; (iii) even if te tion of project-based povery alvlat i,m- direction of an effect is predicted cleary by ecoures ("social ac programe); and echil as nomic fteoy, te mnwditude of de effect is an sistance, tg ad Institutionilding in emp issue and may be a nmter of conto support of the foreing ents The agenda ve; and (iv) economic theory is more useful of the SPA Program is, there an operal for th comparive staics anysis of dieet one, includingdhld4evlnderin and equitm staes tha it is for dte 9tdy of tanthe improvement of the institutonal, policy sition psses and the tiu lags involved. managemnt and krowledge base for povery Therefore, and in view of the goal of the reduction. -SDA Progam to contriute toward improved, The objective of, the SDA Prm to en- poveconclous m and sectoral hance temkowledgease for prty eduction policy magmen, it is nea to compile in Sub-Sahaa Aca equies appled policy stateof-ea alns todealwith anayss and te dffcent coon of prmrt eidentlfleddlffclesande ued data However, te policy res pm at reseatch and economic and sctr work on the hand (i., e assessment of th employment, social dimnsions of adjusmnt and, indeed, of )~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

11 developnent. Ths is the objective of the Volume teir members; and the Community Sumvy, deon SDA Aralysis Plans The AJaysis Plan are signed to collect infmation on the social and detailed conceptual and methodological treatises economic infrastructure and the functioning of on how to analyze key sodal dimensions on the factrand product mukes, induding prices. The basis of cross-section household survey data. Priority Survey can be used as a monitoring in- They are meant to guide researchers in countries strument. It is designed for rapid enumeration participating in the SDA Progm, and they for- and proceing in order to generate quick results. the basis for training materials to be developed as An Analysis Plan for the SDA Priority Survey is part of thecapacity-building activities undertaken included in this Volume (Chapter Ten). by the Program. Social dimensions of adjustment can be Analyzing the meso-econonic effects of identified along three broad axes, viz., the markets structual adjustment in which the poor trade, the human resource sectors, and the livelihood conditions of selected socdoeconomne,grups likely to harbor a particu- Adjustment policies can be expected to influence markets and infrastructure through the larly large fraction of people below the poverty following major effects: (i) relative price changes line. In line with this notion of the three axes, the as between tradables and nontradables; (ii) price Analysis Plans assembled in this Volume focus changes within these two broad categories of on the nature and extent and (to a degree) causa- goods - for example, trade liberalization can be tion of poverty, labor markets and food markets, expected to reduce the price of previously prohuman resource effects of structural adjustment tected importables relative to exportables and in health and education, and the impact of ad- other inportables; (iii) various specific pice polijustment on women and smalwholders. A two- cies (involving changes in consumer taxes and step analytical approach is posited, viz., macro- subsidies) which will furh modify the relative meso analysis to determine the effects of macro- price changes induced through the real exhange econondc policy reform on the weso-economy rate depredation and trade liberalization; and (markets and infrastructure), and rieso-micro (iv) changes in public expenditures. Insight into analysis to trce household-level effects and re- how an adjustment program has affected the main sponses to meso changes An Analysis Plan on marketsand, through publicexpenditurerefonn, the meso-economic effects of stuctural adjust- economic and social infrastructure, cannot be ment is therefore presented at the outmel As part achieved for evety specific market or expendiof the SDA analysis package, it serves as an es- ture t rogram- there must be some aggreation. sential input into the other Analysis Pian3 in- rwereastronggaly naydempiicalgrounds cluded in the Volure. The "macro-meo Chapter seeks to explain how some of the economic for aggregating products according to their tradability in world markets. The Analysis Plan variables taken as explanatory variables in the discusses the issues raised in making these agoer Chaptes of the Volume are themselves in- gregations. fluencd by policy interventions. The Analysis Plan then explains the proc- The data source assumed to be available ess of adjustment (expenditure switching and to apply the Analysis Plans is a nulti-topic cross- expenditure reduction) going back to basic section household survey, complemented, where macoeoomicidentities. Methods of estimating needed, by information from a onumunity sur- changes in the tradable/nontradable price ratio vey on prices and the availability of sodal and (the real exchange rate) over time are proposed, economic in*rsructure. The SDA Progrm has includinganempiricalexamplefrommlakwiand worked to develop three survey types for use im participating countries, depending on analytical Tanzania Expenditure-reducing effects are ana- lyzed by identifying which of the broad categopriorities, resours and "absorption" capacity, ries of expendibe were subject to cuts (this inviz., the Prioity Survey, designed to identify tar- volves the construction of tm series showing get groups for social aspects of adjustment pro- the changes in cental govement public spendgamns and establish for such groups key indica- ing by major categoy). After the exposition of tors on welfare and basic needs fulfillment; the lntegtr Survey, designed to collect comprehenanalytical approaches relying on the historical record, theanalysis Plandiscusseseconomy-wide sive data on expenditures, income, assets and socio-eoanomic characteristics of households and nmdelling techniques. It suggests a series of feia tures which such mnodels should possess. indud- 2

12 ingpolicyrelevance, amnlmumdegreeofdisag- ductivity Improemnts through highr investgredaon, a capbility of representi nucro- nents are therore essewtl if both employmet poucy Intrmns nd a capability of treating and eamings are to rise togetwer. This Analysis pvoduct and facto prices as endogenous. Plan reviews a number of different elements and suggested aoache for tracing how employ- The poverty pile ment and eamings change over time asa result of the policy effects of adjustment. One of the first questions addressed In The Chapter begins by oudining the imthis Analysis Plan is the concept of poverty that is pact of adjustnent on employment and earnings to be used. Recent work on poverty has tended through product-, labor- and capital-market eftdadoptrelativepovertyconcepts,pardybecause fects. It then suggests basic crowtabulations this avoids the need to define some absolute which should be developed for previously deterpoverty line. The Analysis Plan suggests that in mined categories of socio economic groups. For cases where a poverty line is kmwn and is ac- employment, basic data would include informacepted and used by policy nakers, this should be tion on labor force participation, employment appled to measur of poverty. However, should charactdrsti of wage es, the selfemployed, no such threshold be in common use, analyses the formal/infonnal sector, and unemployment based on relative poverty should be undertaken. and underemployment. For earnings, basic data A review is made of alternative indicators would include the distribution of earnings, Of household and individual welfare which would sources of income, the nature of non-farm houseform the basis of the analysis of poverty. Both hold enterprise income, charactics of houseeconomic theory and practical considerations holds receiving minimum wage income and insuggest that total household expenditures per dices of real eamings. capita (or per adult-equivalent) should be taken The Chapter then outlines methods of as the basic welfare indicator in the poverty pro- multivariate analysis to determine wage earnfle, though the Analysis Plan does List a number ings, including earnings functions and associated of altemative measure_ estimation problems, as well as methods to cor- To reflect the intensity and incidence of ree for selectivity bias Methtods to nalyze wage poverty,aswellasthe degree of inequality among differentials by gender, formal/informal, pubthe poor, the Chapter recommends the adoption lic/private and traded/nontraded activities are of the P Alpha dass of indices. These indices are proposed. Fmally, the specification and estimasub-group decomposable, which makes them tionof profithfctionstoassessthedemnants particularly useful in assessing how the struc- of theearningsofann-farmhouseholdenterprises tural changes associated with adjustment are arediscussed. likely to affect poverty. Indeed, the Plan recummends that overall poverty be decomposed by Analyzlr.g human resource effectu Health socio-economic group. If these groups are well chosen, they should reflect the different ways in Structural adjustment may influence which adjustmnt affects various groups of soci- people's health status by affecting expenditures ety. By identifying which groups are most prone on health-related services and nmrket goods, by to poverty, some assessment can be made of how changing the cost of these services and by reducthe sectoral shifts arising from adjustment inter- ing real incomes due to changes in wage eamings ventions are likely to affect the distribution of and farm profits. But health is multi-dimensional poverty in the country cnned and difficult to measure. It is suggested in the Analysis Plan that anthroponetric measures are Employment and eanings effects useful indicators of health but drawing condusions regarding health on the basis of nutrient Insing employment and engs is intake is considered to be fraught with ambiguicrucial in Sub-Saharan Africa U the depth of pov- ties. (Nutrients are important inputs in the proerty in the Region is to be reduced. Achieving ductionof health but unsatisfatory indcatorsof this objective is dauntig given a labor force health status). The Analysis Plan focuses on regrowth of 2.7% per year, the highest of any devel- duced-form demand equations for health, using oping region. Growth through increasing effi- prices of food and measures of accessibility of denc in the use of existing resources, and pro- 3

13 publc health progra household assew and are identwed. ITey indude assets and housewages and houwehld T pubhe goods (soures of hold c prkesand i re. The drinking water, sanituy hfatie, et.) as inde- bsic research str proposed in four steps, pendent variables. The poicy dsussion devel. viz., (i) pearn of data from te household oped in the plan is in terms of food prkces and sur and oiher relevant soures discussed in "prorms, i.e, health facilites, fertility control the Chapter, (I) estimatlon of reduced-form relaprograms and household "publice goods tons of the type Just described, for all of the rlevarn outcomes; (iii) estmtion of Xt changes in the right hand side meso-prce and infrastruc- Analyzing human resource effects: Education ture varibles that we due to the adjustment policy for a given duration of that poicy by reating The output of the education system in a the charnes in th vaiables used to estimate th country is the result of supply and demand fact alternative versions of the basic reduced-form tom The supply side consists of the schoob and equation to the esimates of the chngs in the institutions, as wel as the teachi body and meso-setting and (iv) use of a different form to materials. Structura adjustment may imply esimate the impte of adjstment policy on varibudget cuts for the public education system in ous outcomes rebated to the detemination of creases in user chares or education, changes in education. pricng policies for school materbls and inaeased rationing of access to public schools. On the Household food secuity and nutrition demand side, adjstnent policies will affect household incomes as well as prices, which will This Analys Plan has the obje*tive of have an impact on the amount of education serv- demontatig how poliy changes introduced ices demanded. during adstment affect food security at the Analytically, the income effect is to be dis- husehold and individua levels and of showing nguised from the price effect The forme op- how these effects can be estimated on the basis of erates though changes in incoe which directly a multitopic household survey. There are at affect the demand for education. The price effect least four specific uses for poly makers from an works through making sdwols drectly more ex- analyss of this type, viz., (i) the asment of the pensive (use fees) and/or thoug changes in Impat of past policy chans on food security; the opportunity cost of attending school The (ii) monitoring of te impact of policy reform on Analysis Plan fos on the detenrmnants of the food security; (ii) the prediction of the impact of dmand for schooling. Three policy isues are policy refbrm on the food security of different distnguished: {i) How much will the ability of groups of houolds; and (iv) the assessment of households to send their children to schlm be the costs and benefits of policy revision or of affected by a reducionin their inconme? (ii) What additional polcies that might be introduced to is the impact from the price effect (ie., the intro- improve food security. These include polces ductionof,ornreaiin,usercharges)?(iii)what that may amelorate the tansitory effects of fte is the long-term impact of adjustment on the de- adjstment process on particular population mandforeducatlon? If adjustenttrulygrates groups growth, Oweeshould bean hnceaeintdedenand The Analysi Plan reviews and proposes for education due to an inrese in the expected methodstoestimate -odexpenditreand returns to education. price-food expenditure elasticities and to use the In theanalysis Pla redued-fomdemmad ruts of ths analyis to identify food-insecure relations ar developed for each of the variables househlds. Crhel attention is paid to the defidetemned by the hsehold, including the time nitnwandwm mn_tof varablsonboth sides uses of all household mem s and including of relevmt demand equation. It then discusses among those time uses the time spent investing applicatons of this analysis in the design of proin education. The neasumrem of appropriate grams to enhe or preserve household food dependent variable (observed in the SDA Inte- security during adpstunt. These indude (i) grated Survey) is discused (this inludes, inter indirect incme tans I the form of general aula, admission to, or eolmnt in, a given type or comnditpecfic food subsidies (ii) use of of school) and the observed p edterine van- public employnut schemes; and (iii) use of tarabses (from th point of view of the household) geted direct income transfers in the forn of food 4

14 aid, cash, food suamps or ration cards entitling women working coniderably lgr hours than the holder to fhe or subsidized food. he Analy- en. Mis gives rise to a classic principal-agent sis Plan then moves on to review nmeods to problem the woman has litde ncentive to work assess vidual-level nutritional status on the well. FinaDly, women bear the burden of reprobasis of ae mea Z-score duction. Because there is a phase during mid-life calculations and multivariate methods are pre- in which women's ime is pre-comnutted, certain sernted. actiities are preduded. The physical deman of child beanng and rearing strain health; studies The impact of adjustment on women show that female health deteriorates relaive to mate health in this period. There are two broad issues that need to be These processes jointly gi% dse to genaddresed under this geneil title. First, there are der-differentiated economic outcomes. This is question conceming how gender-related con- mnifested by two symptoms. First, resources straints may serve to frustate the objectives of an controled by women (of which labor is the most adjustent progm, because they adversely af- important) tend to be alloated among economic fect labor mobility and flexibility in moving to- activities in a radically different manner from ward tradable activities. If women traditionally male-controlled resouces. Women's labor tends cannot engage in tradable activities, an expa-t to be concentrated in activities whose output is sion of tradable output wiu be far more con- intemationally nontradable and which provide strained than it would be if such gender con- relativelylowincomes. Thissuggeststhatwomen straints did not apply. Second, does adjustment may be more constrained than men in gaining enhance or undermine the role, status and wel- access to higher tum activities. Second, differfare of women? Do the changes induced by ad- ential constaints upon access can be observed by justment, especialy the changes in the structure analyzing the process of the transition of resources of incentives, lead to a breaking down of gender- between sectos Structural adjustment is cenreated constants? trally about there-allocation of resources between At the micro-level, economic decision- activities which should expand during adjustmakin is gender-specific because of four gen- ment, then this is liely to constitute a significant der-differentiated processes. Pirst, women may policy problem. The Chapter suggests lines of encounter discrimination outside the household. data analysis (such as Logit or Tobit) which are While in developed countries discriation of- appropriate for multivariate analyses where the ten takes the form of lower wage rates, in Afica dependent variable is discrete. it seems that in the labor market ft more con- The major gender-related task in the mrnly takes the form of differential access to analysis of survey data is to identify the extent to wage employment. Perhaps of greater impor- which policy changes can assist women in overtance is differential access to credit, because coming differential constraints. Typically, access women generaly do not own marketable land to credit makets, education, agricultural inforrights and, as subordinates In the household, mation and land is biased against women. There cannot establish independent reputations for is therefore a case for an offsetting bias in the creditworthiness. Second, imitation, or copying competing provisin of public services. Yet, curis an important way in whih new economic ideas rently it is common for some public services to be are disseminated There is some tendency for biased in the same direction as private provision. men and women to have different role models, The Chapter suggests how survey data can be men copying other men and women copying used toidentifythegender-specificeffectsof som other women An implcation is that, if some important public policies new economic opportunity is initially taken up by men, it may automatically be diffused over Smallholdens and stuctual adjustment the male population but have a slower diffusion among females. Third, win the household there In most economies of Sub-Saharan Africa are asymmetric rights and obugations. Women the overwhelming majority of agricultural proare obliged to grow food for subsistence, to gather ducers, and indeed the vast majority of the total fuel and water, to cook and to rear children. In population, can b- dlassified as smal1holders. At retun, men meet certin cash needs. This pat- the same time, the low or declining levels of agritern of reciproc obligatons is often unequal, cultural productivity in the region are, to a large 5

15 extent, a reflection of the ecownmic and physical Al analysis plan for the SDA piqorty surey conftaints imposed on snaulholder agriculture. This environment is charactized by very lim- The concerns addressed In the previous ited access to the mourms necssary to rise plans require in many cases in-depth aalysis of productivity and by inadequately developed the data which the integrated survey is designed transportation and marketing infrastructure. to provide, in many instances jointly with data Snialiholders frequently operate at close to the from the community survey. It goes without margin of survival and under uncertain cdimtic saying thkt a certain amount of time wiu be i- and market conditions. Tir decsion-miaking, volved to complete these analyses. This is war6 therefore, is distinguished by a conservatism in- ranted in view of the fact that many of the issues dicativeof risk aversion, which nmy circunscribe addressed have a medium to long time horizon. their ability to respond b new opportunities. They wiul serve to put in place policies over the The smaholderanalysisplanbeginswith next three to five yeams and beyond. In many an overview discussion of the way in which cases, they provide an input in development structural adjustment is expected to affect small- strategy as such. holders through exected changes in markets for The priority survey addresses the other products and labor. An uncerainty is expessed side of the time spectrum, the immediate deciregarding the short-run effect of adjswment on sions that need to be made. The policy focus thus rural wages relative to both the cost of laborers shifts from structural elements to symptoms and consumption basket and the pnce of agricultural indicators of immediate need. For exapl, while (tradable) output. A classification of smallholder development and long-term policies nuy be exhouseholds (including infomlal sector non-agri- pected to solve the malnutrition problem, In the cultural households in rurl areas) is proposed, meantime starving children must be fed if they sart with a definition of snallholdersin terms ever are to be the beneficiaries of these longer of the ratio of the amount of labor hired to total term measures. The priority survey therefore has labor used in the hou*ehold agricultural produc- a strong focus on poverty and basic needs betion process (the ratio must not exceed one for cause these are the key areas for immediate atenhouseholds to be classified as smallholders). Classification criteria are the relative importance tion. Since the construction of income and expenditure measures cannot be undertaken of wage to total income, the choice of crops and through this instrnuent, the nain challenge is livestock activities (tradable/nontradable and, in the design of suitable proxy nmasures. Analytithe case of the former, food crops for import- cally the task is to use these measures to identify substitution versus export crops) and locational target groups. criteria. Five classes of agrictural ouseholds The policy focus of the analysis is thus not are proposed on this basis, viz., export-oriented produces, food-crp-oriented producers, wage so much the modification of employment educa- tion or health policy, but rather to identify those laborers, subsistence farmers (i.e, families deiv- groups in society which are particulary disading a particularly low proportion of their income vantaged in any one of the dimensions that confrom wage activities and/or the sale of produce) stitute the standard of living and to document and informal sector rural workes the extent of the disadvantage. The analysis will lteanalysisplanthendiscssmetowds thus not necessardy provide indications about for the classification of acltural commodities according to their degree of tradability, and it the type of needed iterventions to solve the problem in the long term This indeed is the role reviews approaches to the calculation of rural of the sector-specific Analysis Plans. Here the terms of trade and coefficients of net protection aim is to set priorities based on observations of of agricultual output. The famihar agricultural the eent ofa problem in diferent soco-eonomic household model is then intduced as the sug- groups. For exiample, if in certain groups malnugested analytical device to shtuy the simultane- trition is severe, this provides information on the ous detmination of various production and consptionchoices, including the optimization direction of food aid. he actual provision of that aid will require solid knowledge of the country of the combination of off-farm and on-farm employment. and of what type of interventions are feaskible in the short run in the country. 6

16 Introduction and overview TheSDAProgr,laI&un&edjoltybythe he timely deveopmen of opeatonal reci- AftnDevopmentBa,UD Pandthe World mendatons in priority poliy areas) is arguay %aiuk, withhe World Bankas ecutingagency, complex. In a nutshell, the SDA research task is and fuwded by a number of nuldtlateral and bi- to twace mac-micro relationips in order to latera doos, spports parlicipating African understand the processes through whdch the efpwernments (30 to date) in their efforts to assess fects of moeconomic policy reform (or exterand monitor the sctoal and groupspecific con- nal shcks) are transmitted to households. This sequences of omic adsmt The is less than strihtfor because of theoreticontributions of the SDA Project indlude activi- cmlimtations and data constaint at the houseties toward Wm rved, povertyconscous macro- hold level in SubSaharan Africa. Macro- and economic and setoral polcy managemen (pol- miconics lack theoretical Integration, and icy studi~s and the ognization of fora to fili- macro-micro analysis is complicated by the fact tate exhange betwen polic analys and deci- that - as argued in this Volume - (i) private sion-makers) the strengthing of natonal st. behavior may offset the intended effects of adtl systms with a view to geating the justnent policy on prices and the avabity of household and commity-level idormation re- goods; (ii) net market effects may depend on quired by the above; tlhe design and implemeta- ne iresponseswithoyofferinglitde tion of project-based poverty aleviatbn mess gnada reganwh tr relativeimportanc, (ill) ures (social action programs"); and tednical evenif thedretionofaneffectisprediteddearly assistance, training and institution4builing In by economic theoiy, the magnitude of the effect support of the foregoing elements. Tb1 agenda still is an empical issue and may be a matter of of the SDA Pam is, therefore, an operaonal controvr and (iv) economic theory is more one, indludingdrect field-leve intvenions and usful for the comparative statics analysis of difthe Imprvement of the institutional, policy fereitequulriumstatesthanitisforthestudyof manageent and knowledge base for poverty trsition processes and the time lags involved. reduction Therefow, and in view of the objectives of This Volume is concerned with the policy the SDA P ngm nmeoned above, it is necesanalys strategy ofthesda Prgrm The objec- sary to compile state-of-thewrt analycl instrutive of enhancing the knowledge base for pov- ments to deal with the identified difficulties and erty reduction reqires rearh and the effidient guide country-based research and econonic and collection of primay data. However, the polcy sector work on the sodal dimensions of adjust- _search problem at hand I, the assm t of ment and, indeed, of development. This is the the employment, disrbutil and welfare con- objective of the present Volume on SDA Analysis sequences of sbl otbn and adjustent and Plas. 7

17 The Anayis Pla are detd concep- corcmtrtes lrgely on outome variables, can tual and methodological treatses on how to ana- be implemented and analyzed in a matter of ftree lyze key sodal dimensions on the basis of cross to four monuts. Its properties and its relation to section household survey data. The socal di- the comprehensive, multi-topic household surmensions addressed in hids Volume include the vey(thedatabasefor theanalysisplansinchapnature and extent and (to a degree) causation of ter Three to Nine in this Volume) are explained poverty, the impact of adjustment on women and in Chapter One. smalholders, employment and earnings effects TlhisVolumeconsist of tenchapters and a and human resource achievements in the areas of Technical Appendix and is structured as follows: health, education, and food security and nutri- In Chapter One, an analysis strategy for the sotion. The anawysistrategy of the SDA Program dal dimnsions of adjustmt is outlined, focsposits a two-step approach, viz., macro-wmeso ing in particular on methods for macro-micro analysis to determine the efects of macroecon- analysis and impued data needs, and identifying onic poicy refozm on the meso-ecodnmy (nmr- key sodal dimensions and associated polcy iskets and infastructre), and meso-micro analy- sues. Chapter Two, on nucro-meso linkages, sis to trace household-level effects and responses shows how to analyze the me-economic effects to meso changes. The aim of the Analysis Plans of structwal adjustment, ie., the effects on prices is to provide guidance to researchers studying in product and factor markets and the effects on te social dimensions of adjustment. The Analy- economic and social infrastructure working sis Plans are important elements in the set of throughchangesinpublic xpenditure. InChapanalytical tools developed by the SDA Prgram ter Three, a detailed approach to the construction for the purpose of assessing the distrbutional of poverty profiles is presented. The identificaand welfare consequences of adjustment These tion of the poor and their classification in terms tools also include survey methods, and an Analy- of Iniome, expenditure and basic needs characsis Plan for the SDA Priority Survey (a lage- teristics is the logical first step in appled social sample mo toringinstrunment) is included in the dimesi analysis. t is shown that social dicollection in this Volume mensbons can be identified along three axes, viz., Sodal dhnensions analysis cannot be car- marets, sectors and target groups. A key married out in Sub-Saharan Africa wihut primary ket deterdnige income is the labor market, and data codection at the household level. There em t and eanigs effects of structura exists considerable experience with household adjustment and the functioning of labor markets budget surveys in developing countries. Field- are analyzed In Chapter Four. In Chapters Five, work for these surveys is normally spread over a Six and Seven sector effects in the area of basic one-year period and the analysis can be tine- needs are analyzed, fcusing on health, educaconsmun, even though there has been great tion, and food security and nutritionl In Chapter progess in the past 10 years in data processing Eight, the impact of adjustment on a major taget technology. TheSDAProgramhasdevelopedan group, women, is analyzed. Chapter Nine is integrated household survey questonnaire de- devoted to an analysis of smallbolders under signed to pemit extensive multivariate analysis. adjusment. Chapter Eight, on women, and In the process of this work, however, It becme Chapter Nine, on snumlholders, overlap in part dear that (because policy action cannot await the In Chapter Ten, finally, the analytical possibilresults of extensive surveying and analysi) a tiesofthesdaprioritysurvey aresetout. In the lean, much more rapid and yet sysmatic sur- Techdncal Appenix, concepts and tchniques to vey instument was also needed. The SDA Prior- calculateincomeandexpenditureaggregatesfiom ity Survey is the result of innovative methodo- the SDA ntegrated Survey (see Chapter One) are logical work to balance the sample size Implica- pested. Making these techniques (and contions of the need to obtain stistcaly represen- ventlons) explidt is of overriding importance in tative esfimates from heterogenous popuations, the efort to develop consistent data sets for use the need for short interview times and the re- by wcesive, Independent groups of researchquiemnent of speed in processing and tabulation. es. It is estimated that the Priority Survey, which 8

18 ~~1 An analysis strategy for the social dimensions of adjustment 1.01 In settng out an analysis strategy for the work suffidently wer to permit the intended resoial dimnionsox of adustrent it is necessary to allocation of productive resources from nonexplain how mceonomic policy reform is tadables tradables If sucessful, adjustent linked to micdoeconomic outcomes and wefre will lead to renewed growth on the basis of doconsequences (paragraphs 1.02 through 1.11), to mestic rather than lgy imported resources, identify thekysociadlnensionstobeaddressed although the suctural import dependency of (paragraphs 1.12 through 1.16) and to identify African economies may hamper this process. the data requirements and specify an approach External mbalane wil be corected under subto the colection of primay data (paragaphs 1.17 cessful adjustment due to the revival of producthrough 1.24). Key poliy concerns assodated tion for epo and efficent import substitution. with each of the social dimensions analyzed in Adjustment is expect to incrase equity by inthis Volume and an analysis strategy statement dudng a rise in the share of labor in the overall we the subject of paragraphs 1.25 through 154. factor combination in accordance with the compaative advantage of the African ecoomi un- Adjustment and distribution Considerations der study. towad an analytical approach 1403 The real exchange rate is influenced, bmadly, by policy measures affecting the normi Structural adjustment seeks to restore nal exchange rate, moneary and fiscal balances interal and ar te balance and hreby create anud the regime of trade protection (Net capital anew basis for sustnable economicegrowth with infows are a further determinant of the real exequity. Adjustent programs consist of nmas- change rate.) Ihlerent in tese policies are eleures of price policy, broadly defined, whic! are ments of both demand reduction and expendioften accompanied by institutioral changes to ture swithing. Whs monetary and fiscal Improve the functioning of markets and the effi- contraction can be recessonary (but beneficial cency of the pubwc sector. Itis useful to view the from the point of view of inflation control), the price policy dimension of structral adjustment restruc of taxation and publicexpenditures as the opeing of the national econmy by mens under adjument can and must be designed to of the deprecation of the real exangw rate and support growth through more effective, less dishence the improvement of the price of tradables crininatory taxation, a higher-yielding public relative to the price of nontradables. The devalu- invest portfolio, human capital formation ationoftherealexdhangeratelwadstodhangesin and adequate and institutional the disbuttion of incentives and income and support to the tadale sector. Similarly, trade cam twfore, introduce effctve structural lberalization (i., more neutral pricing and dhange in the economy, provided that markets greater openneses) is likely to contribute to 9

19 anstory unemployment in some sectors, but is expected, in timne, t enhance prduction and inurem to address the sodal cost of adjustment by raisingfactormobility,creatingemploymentand come possibilites in the tradable sector which, improving food security and access to basic servtypically, is "home" to the majority of the popu- ices? This question is closely tied to that of raslation in Africa, assuming, as seems corect, that ing the poverty focus of public expenditures in the tradable sector covers much of agriculture as the context of fiscal refomn, by ensuring adequate well as import-substituting informal sector ac- funding and effective program delivery in areas tivities. (Criteria for the classification of goods of vital concern to the poor, viz., primary social and services into tradables and nontradables are services, food-linked transfer payments and deoffered in Chapter Two.) centralized economic infrastructure and services While prompted by financal imbalances, The Analysis Plans on basic needs achievements adjustment thus brings about changes in the real in this Volume provide detailed discussions of economy. In the standard "dependent-eonomy" these policy issues. model of adjustment it is assumed that real de In theattempt to shed lighton these policy valuation is possible and will have an expansion- Issues, the SDA research task requires the tracing ary effect. There is, indeed, positive evidence of niacro-miro relationships in order to underregarding the possibility of achieving real de- stand the processes through which the effects of valuation- nder African conditions(world Bank/ macroeconomic policy refonn are trasitted to UNDP, 1989), but the issue requires constant households Thisisbecausewelfareisamatterof monitoring in view of the possibility that in situ- households and individuals, i.e., it is a micro ations characterized by serious balance of pay- concern However, tracing macro-micro relanents difficulties devaluation may work at best tionships is less than straightforward because of as "a tool that validates past monetay expa- the well-known lack of functional, theoretical insione (Kateeli, 1983: 360). tegrat{on between macro- and nmcroeconomics From a social dimensions point of view "Microeconomics focuses on the decion-maktwo sets of research questions arise from the above ing behavior of individual units, in a partial equicharacterization of the adjustment process. First librium, ceteris panbus world.... Macroeconomis a set of questions dealing with the distnbu- ics, on the other hand, focuses on aggregate ecotional effects of the relative price changes trig- nomic constructs and takes feedbacks into acgered by real exchange devaluation. In particu- count, but is unable to relate the overall changes lar, how does adjustment affect production, em- with which it deals to what is happening to indiployment and consumption possibilities of the vidual units. The 'aggregation problem' is usupoor? (Consumers of nontradables and produc- ally swept under the rug by assuming that aggreers of tradables are expected to bo affected favorably relative to consumers of tradables and gate changeflects microbehavior, without con- sidering the role of compositional or structural producers of nontradables.) How do changes in change" (Ruggles and Ruggles, 1986: 245). A key the identified policy detenrinants of the real ex- challenge for SDA researh clearly concerns the change rate affect specified socio-economic establishment of links between "overall changes groups? And what are the timepaths along which and 'individual units. The feasibility of obtainthese policy changes act, noting that demand re- ing sufficiently robust answers, for policy and duction measures tend to be effective in the short- program purposes, to the question of how adjustrun, whereas restructuring policies may be slow ment policies affect poor and vulnerable groups, in taldng effect. therefore, is a major issue in SDA policy analysis A second, more action-oriented set of re- A two.stge methodology is proposed, i.e., macrosearch questions is as folbws: Are there alterna- meso analysis to uncover the effects of policy on tive sequences and combinations of macroecon- factor and product markets and economic and omic and market-level polides that entail smaller socal infrastrucmte (the meso economy); and welfare losses on the part of the poor without meso-micro analysis to illuminate the consejeopardizing the objective of restoring macro- quences for household welfare. economic balance? What meso- and micro-level 1.08 The interactions between the two analysis policies and programs are needed to enable pro- stages (macro-meso and meso-micro) are disducers and consumers to take advantage of the cussed and formalized in Chapter Two. Mesoopportunities offered by changed relative prices? micro analysis is the study of incomegenweration What are appropriate interim or short-run meas- processes, expenditures and welfare indicators 10

20 (for example, nutritional status indicators, edu- onnmaintenancespendingmayhamperthisproccatonal achevements) as a function of prices in ess, but this depends in good measure on the the markets with which the household interacts, pattemn of aggregate demand reduction chosen householdassets and characteristics and relevant by the government and associated resource iminfrastructure available to the household. Stan- plications for infrastructure considered essential dard utility maximization theory provides the to obtain a supply response and provide essenframework for this analysis. The most pragmatic tial services. This kind of infrastructure spendway is not to estimate outcomes from the data in ing, as weu as subsidies to the consumption by a structural form but through reduced-form de- the poor (where the poor are effectively reached) mand equations for relevant household choices. should not be cut in the context of fiscal reform. (The reduced-form demand equation for health Instead, cuts might focus on non-performing deis formally derived in Chapter Five.) Key areas velopment projects, defense and consumption by of 'choice" from the point of view of social di- the rich. While these choices are inevitably pomensions are those related to the satisfaction of fitical,thereareanalyticalissuesintermsof macobasic needs (housing, food security, primary meso-micro analysis which must be investigated. health care and primary education). Reduced- What are the implications for growth, participaform equations can thenbe used to simulate what tion of the poor in economic recovery and basic would be the impact of various elements of needs satisfaction of alternative patterns of agmacroeconomic policy change on relevant gregate demand reduction? household outcomes (see the four-step research 1.11 Afurtherthemeofmacro-mesoanalysisis strategy in Chapter Six). Adjustment changes the that of the best combination of macro policy inpnce and (through changes in public expendi- stmments (notably exchange rate and fiscal and tures) in&fastructure variables in the reduced- monetary policy) from a poverty reduction point form This, then, will lead to changes in the of view. While individual lines of expenditure outcome (or dependent) variables of interest. may have to increase to achieve goals of recovery 1.09 The Analysis Plans presented in this Vol- and redistribution, total expenditure may need ume (except Chapter Two) concentrate on meso- to be cut in the early phase of adjustnent (stabilinicro linkages. Both descriptive analysis (cross- zation). For given levels of external finance, the tabulas)andregression tehnqueaisingfrom extent of fiscal retrenchment and the length of household moedd structures (unspecified in most the period of austerity depend on the speed at Chapters) are presented. In this approach, rela- which growth resumes. This, in turn, may be tive prices and infrastructural factors are treated influenced by altemative rates and sequences of as exognous variables. Macro-meso analysis, in exchange rate adjustment, money expansion and tur is about how the price and infrastucture deficit reduction. For example, whether a conchanges occur. This, and problems encountered traction in money supply leads to output reducin the weasum t of product price changes tion is likely to be influenced by the rate of deand changes in rea wages, credit markets and valuation. Similarly, if expenditure cuts in key public expenditures is discussed at some length areas for growth lead to a decreases in output, in Chapter Two (paragraphs 2.36 through 2.67). the maoeconomic constraints will tighten be- If data on price and infrastructure changes exist cause of a reduction in tax revenue and/or the or can be constructed for a specified time period, country's creditworthiness and ability to borrow. the resulting hanges In dependent variables (or The chaulenge is to find the combination of exoutcomes) can be estimated in the reduced-form penditure cuts, revenue growth and monetary framework, as explained in Chapter Two. In this and exchange rate powy which decreases the approach, however, the determination of housew deficit with the least adverse effect on supply. hold income is not linked to macro policy change Thus, alternative macro strategies have different and prices are still exogenous. Toovercomethese implications for growth and redistribution, and analytical constraints it is necesay to build a these can be traced, in principle, by means of geneal equilibrium model of the economy, and economy-wide models. this, again, is discussed in Chapter Two It is suggested in Chapter Two that infra- Identifying the key social dimensions structue, Like pries, plays a role as a conduit to pass on adjustment-related signals to investors, 1.12 Having introduced our two-step analytiproducers and consumes. Resource constraints cal approach, it is now necessary to propose a 11

21 way of organing the multidimensionality of gated - the task of the poverty profle.) The poverty and the social dinsions Sodal di- selecion of smallwholders folows from the fact mensions are appropriately defined along thuee that in African countries the majority of the axes, viz., markets, sectrs and target groups. population is rural, and agriculture is the domi- The key markets in which the poor trade and nant sector in the generation of the countriese whose prices determne real incomes are labor national product. The bulk of agricultural acmarkets,creditmarketsand themarketsforgoods tivities takes place on small farms. The andservcesproducedandconmedbythepoor. Smialholder Analysis Plan, while focusing on a Among the latter, food markets ame of particular taet group, is also parallel with the employment importance, given the role of food as a major plan in that it foces on the income-generation wage good and input in the production of health mechnhsms of this particular group. The emand nutritional well-eing. Food markets and ploynent plan focuses on urban areas and the household demand for food are analyzed in smallholder plan on rural areas. ChapterSevenbelow. (Promtheproduceespoint 1.15 The selection of women as a taret group of view food markets are analyzed in the Analy- follows from the fact that in generating income sis Plan on smallholders, Chapter Nine). On the men and women do not face the same constraints incomw side, labor markets and employment are in taking advantage of income and employment the primazy area of study. The Analysis Plan on opportunites Their control over resources and employment and eanings (Chapter Four) ex- their ability t spend income on consumption plansthemxchanismsthghwhichhouseholds differ Siinlarly, their demands for social servgenerate income. Emplyymentisnotconstrained ices (i.e., the sectoral demands) are typically quite to wage employment, but includes the genera- different. The corresponding Analysis Plan is tion of household enterpriselncomeaswell. Plans not meant to suggest independent analysis for sinilar to the employment and eamings plan the case of women, but rather to demonstrate the could be designed pertaining to the credit market importance of providing a gender breakdown in and selected product markets. lhese would defi- social dimensions analysis and to illustrate how nitely be options to be considered for specific this can be done in important areas such as countries. women's participation in the labor market and 1.13 Along the sectoral axis, social dimensions education. This Analysis Plan makes the general resarch is particularly concened with human point that households cannot be the final level of resources, i.e., education, health and food secu- analysis in loolkng at the social dimensions of nity and nutrition. Health and nutrition are di- adjustment, but that at least the gender distincrectly related to the survival of individuals, while tion needs to be made at the Individual level. education is the main element of capital acquisi A complete body of social dimensions tion that holds a promise for the fture escape analysis would cover all relevant markets, out of poverty. Each of these sectors represents "achievmnt sectors and target groups. The crucial welfae-related outcome vatiables at the prioritization of these sectors and groups for sohousehold and individual levels. Each sector cial dimensions analysis is a country-specific also is associated with important infrastucture matter. For the puposes of the SDA Program determinants at the meso level (for example, thosesectrs/targt gups which are deemed to school and health failities). be of prime importnce under adj-ustment and 1.14 Finally, social dinsion policy research which are believed to be of relevance to the mamust be target group-oriented. The idenfica- jority of countries participating in the Program tion of specific target groups of interest will be were selected. They becane the topics of the much more a county-specific natter than the Analysis Plansassmbled in this Volume. identification of sectors. Neverlss, it is felt that there are at least three broadly defined tarwt Data soures groups which must be considered in any analysis of the impact of adjustnent, namely, the (pre Infmonation is required at each of the adstmnt and post-adjustet) poor in gen- macro, meso- and micro-levels if policy makers eral, wonmn and smaliholders. The reason for are to know what changes are affecting housethe selection of the poor (idend through the holds and how these chnges occur. It is therepoverty profile) is self-evident. (For poicy pur- fore useful to view infornation requimenst in poses, this target group needs to be disaggre- the context of an hierarchical information system 12

22 where data and empirical analysis are explicitly dertaken. considered at each of the three levels A dlstn- (1) The Integrate Survey (IS) is designed to coltion can bemade betwen "aralysis,""constructs lect, for a relatively restricted sample, compreand "data." Thus, analysis can be either formal hensive data on income and expenditure promodels or anaytical studies. Thse analyses de- cesses, assets, soco-economicharacteristics and pend on empirical infomation, and a disdnction welfare indicators of households and their menis made between the raw data sources (cnsses, bers. This provides baseline information on the suveys) and the various constructs used to as- social and lhingeonditionsof different groups in semble, organize or present these data for ana- society. The survey will pert in-depth analysis lytical use (national accounts, tabulation, sum- of household responses to changes in their living mary statistics or indicators). and worldng environment, which is essential for 1.18 An hierarchical information system for nmdium- and long term policy planning, includsocial dimensions research is presented in Table ing the design of adjustment policy with a solid 1.1. The reader is referred to a companion vol- focus on social concerns. ume, World Bank (1990), for an explanation of (Oi) The Commuity Survey (CS) collects informathe rationale underlymg this system While some tion on the social and econonmc infrastructure data on maco, meso and, indeed, micro vari- and on the functioning of factor and product ables are already collected in most countries markets (including prices). These elements con- (though they are of variable quality), household stitute the meso-economic environment - the survey data are the most critically deficient infor- intermediate link In the macro-micro transrnismation source currently hampering social di- sion chain of adjustment - within which housemensions policy research. A key contribution of holds and individuals behave as economic and the SDA Program is, therefore, to improve the sodal agents. The CS can be conducted as a qualityandavailabityofmcro (andmeso)level separate survey or tied in with the IS or PS. In data though household and conmunuitysurveys. many of the Analysis Plans presented in this This is the basic data source required for meso- Volume it is concluded that comnmuity-level innicro analysis of the kind proposed in the Analy- formation is desirable to complementhe dmndsis Plans below. side information available in the IS with indica The SDA Progam has worked to develop tions from a CS regarding the supply of, for exthree separate but interlinked prototype surveys ample, health facilities, schools and communicawhich could be used to strengthen national infor- tions infrastructure. The CS is a potentialy immation systems. These are portant data source in analyses of the costs and (i) The PrUr{y Sumey (PS) aims td identify taget benefits of alternative urnal resource allocation groupsforsocialaspectsofadjustmentprogamts choices in public expenditures. It permits the and estabish for such groups key indicators on establishment of a link between public expendiwelfare and basic needs fulfillmenl Compared ture data and program performance at the comwith the Integrated Survey (below), the content munity level. of the PS is much more restricted, but the data are 120 The AnalysisPlanspresentedinthisdocucollected over a larger sample and penrmt much meat make heavy demands for micro-level data faster tabulation and analysis. Ideally, the PS is collected from household surveys. Although a conducted as the first component of a survey great deal of analytical work can be undertaken program, hence providing policy makers with an using single subect surveys, the issue of measurinitial empirical basis for targeting essential and ing and understanding the standards of living of urgent Interventions. When repeated, the PS also different household groups can never be adebeodmes a monitoring tool for tacing changes in quately dealt with undess there is the possibility wre and basic needs fuliment across target of corelating economic and social data at the groups. Repetitions of the PS are recommended, household level. Thus a key requirement of the because themncro-leveeffectsof macroeconomic Analysis Plans is that they be able to draw on policy reform are distributed in time and thus intatedmulti-subjecthousehold data sets. The call fora combination of cross-sectional and over- main source of such infonration is expected to be time data. The PS represents a smplfd method- gernerated out of the IS. IS-pe surveys should ologsyforandays, developed by thesda Program be viewed as strategic inmements to generate to yield useful results rapidly, before the in-depth the data needed for longer-term policy planing analysis based on the Integrated Survey is un- and development strategychoice. They take time 13

23 Table 1L: A hieraral bmaton system fo the SDA (# O) (4) (5) () SMA Fows Ass cs*bwo D Mactoa Policy 'Monetary, ftscl.macroaaelc 'National * Enomk *onurs and echaerate mode: 'SAMs tode, ponicy -AMvea1(MSM) *CFI fnanda -Multsectoal (CGI *B3OP stalss *Macro4ndicators Socdal and demopqb Meso Ects 'Markets *Sd and SAMs * COMMUnIty * Economic and institutional studies * Food balae sheets surve sda *CGE *Sectorequantty *Pd rmunszmket modes arkdprinhlce staics Prooducn sbitics *lidividua ohouehold modls 'Povety proiles 'iousebold and household and analyss Soso" Indticalors surveys: $Smai smucluna4snnntd WoMknk Foerty: ACfoU.tIUSVhtJJikS ad l*yinrewmkr.sdau.amcstfreg(otmreportnlo.8&ap&, O;.B, to implemnt (data couecion itself covers a full distutional changes and thus limits their useye) and analyze. Therfor, short-term policy fulness relative to survey-basede action, by defirition, cannot be guided by IS- Furthermoe, even up-to-date consumer price based conclusionsand reo endations. Tis Indies for capital citiesare are in Affica (eferfunction can be fulfflled by. te PS, which has ence consunw baskets are often outdated bebe deigned rectly for rapid enumation of cause of the relative prke changes known to have welfare h st under the SDA Program. omrred during te 198) There is thus merit An Analysis Plan for fte PS is included below as in a rapid, sample-based assess_mt of employ- Chapter Ten. A fuh description of the innovative nient, wage rates, income sources, expenditure PS, nicludig a prototype qetdonnaie, is avail- proxies and basic needs indicators such as can be able in a forthcoming volume, World Bank obtned by means of the PS, which can be con- (199W. ducted over a two-month field period (to whdch 121 As an inshtrumento study changes in proceslng andanalysisetinofanothermonthor poverty and basic needs adcevent the PS is two must be dded), with individual interviews situated somewhere betwen indicator-bad in tng less than one hour (see Chapter Ten). vestigation and the IS. Food price indices, the 1.22 The naret-spefic, secbral and groupreal wage (particularly fir unsled labor) and speific Analysis Plans presented in this Vohme the runrl terms of trade are useful indicators to are "written for' the SDA Integrated Survey and studychbangesinthelving staxardsof thepoor. complemnted by a Community Survey, where In fact, socd dimensio analysis should staitby needed. The Analysis Plans thus aim to expwlain looking at these variables where observations on the concepts and *etods needed to try out them exist or can be ostucted from publihed detaied, causal" policy analrysis (in a houseprice data. Out indicators are not normaly avail- holdlevel, as opposed to an economy-wide, able lfr regional and soo-ecoomic dis - sense)onthewrioussodaldimenionsidenftfied tions, which makes them incapable of tint above-anendeavourwhihcannotbebasedon 14

24 the PS, because thds survey lends itself only to bemetbeforea psonlsconslderedamemnberof lihdted multivarlate analysis. An IS draft protow the household. The defiidtion of the househd typeqeiomaire maybeobtned from thesda used in the IS i a dejure one, but one whih dif- Unit and wil be available as an annotated ques- fers somnewhat from the de jure definition used in tionnare in a forthcoming report, World Bank other surveys that are concerned with a house- (1990b).(Space linitations predude the repro- hold composition at a spedfic moment In time. ductionof thelsand PSprototypequestionnaires This is because the recall period for a number of in this Volume.) The IS prototype questionnaire IS questions is 12 months, Therefore, a defition is divided into 12 modules, including a house- of the household which is valid for this full pehold roster and sections on education, health, rod must be selected. In reality, of course, employment, migration, housing, agriculture, household composition can change substantially expenditures, non-farm enterprises, assets and throughout the course of the 12 months. It is savings and andhopometics. A clarification re- therefore necessay for the IS to use a defintion garding the sectin on agriculture may beinorder. which describes the average household composi- The IS is not an agricdtmal (or farm management) tion during the period. surveyandthereforecannotcolect detailed,crop Three criteria are used in classifying specific information of the type required to esti- whether persons are to be inluded in the housemate production functions. It is similar to the hold or not. First, and consstent with the United World Bas Uving Standards Surveys in that it Nations definition, they should meet the critesacrifices detail regarding agricultural behavior tion of usually living and eating together in the "in order to obtain the broader picture that can dwelling. (The definition of the dweling is a only come from the collection of a great deal of country-specific matter.) The second criterion is interrlated dataonallaspectsof thehousehold's that au of thes persons living and eating toeconomic activities" (Deaton and Benjamin, gether should acknowledge the authority of a 1988:5). Inthesection on agriculture, infonnation single head of household, regardless of whether Is sought on incone from cropping and livestock the latter i living with the other household activities, expenditures on agreultual inputs, mnemers or living away. The situation of poco nsuption of own produce and assets relevant lygamous households can present problems detohouseholdfarmproduction(inparticular,land). pending on whether each wife is treated as a Questions are asked on assets (and, livestock, separate household or as part of one large houseequipment), crops grown, purchase of inputs, hold. In the former case, it wiu be necessary to credit and agricultural processng activities. It apply an arbitrary rule, such as linking the head will become apparent from the Analysis Plans of household to the first wife in order to avoid below that,eve though thissurveyoptionlimhts double counts. The third eement is the time the scope of agricultural production analysis, It is period over which the first two criteria must have able to sustain in-depth social dimensions analy- been met. Itis that peasons should sis for the rural population. have been present in the household for a period 1.23 In preparing for and processing the SDA of atleastsixmonthsoutofthelast 12. Itshould, Integrated Survey, two issues desere specidal at- however, be noted that, beyond these genl tention, viz., the definition of the household as indications, the deflnir'on of the household in the basic unit of observation and the concepts, any particular integrated survey must take into methods and conventions used to calculate in- account local conditions and knowledge regardcomeandexpenditureaggregateswithina system ing fanily structure, mobility and other factors of household accounts. Both of these topics are impinging on the household concept. discussed in detail in the Technical Appendix to this Volume. We shall linit our commnents here An analysis stategy for the social dimensions to the defindtion of the household for IS pur- of adjustment poses. Both de jure and de jkcto definifions are in useforhouseholdsurveys. Thedefixtodefindtion 125 This Vohlme Is about how to assess the pertins to those household members present at sodal dhniensis of adjustment in countries unthe noment the interview takes place and is most dergoing processes of mnacoeconomic policy recommonly used with censuses or other enu- form and/or spontaneous, uncontroled adjustmerations over short periods. The dejure defini- ment in response to changed external conditions. tion relies on a criterion of presence which must The analysis strategy of the SDA Progam con- 15-

25 sists of support to data eldection efforts at the on the direcdon of food aid. The defirdntm of the household and conmumnity level as disussed in actual modalities of providing food aid will rethe previous section and the appication of te quiresolidknowledgeofthecountryandofwhat analytical conepts and methds along the macro- type of interventions awe feasible and effective in meso-micro line discussed in this Volume. The the short-nm precise nature and sequence of data collection and analysis efforts is a country-specific matter Anlysis of the Integrated Survey and cannot be prejudged. In this section, a possible strategy for the asssmet of the social 128 Once lntegwted Sumgy data are available, dimensions of adjustnent is set out, and impor- the analysis suggested in Chapters Two to Nine tant policy issues to be addressed are summa- in this Volume beodmes possible and should be rized. These policy issues are derived from the undertaken in accordance with priorities defined Analysis Plans in this Volume. It is recgnized at the country level At the macro-mao level, this that priority subjects for policy analysis vary would include, ideally, general equilibrium among countries Analysts, therefore, will define analysis of the kind spelled out (but not formalresearch priorities in accordance with local pol- ized) in Chapter Two In order to develop scenaricy circumstances and priorities of policy makers los of the growth and poverly-reduction implica- Nevertheless, the summary of policy issues tions of alternative macro strategies. emerging from the Analysis Plans and presented 129 Regarding meso-micro analysisreseachrs in this section may be useful as a guide in plan- are expected to be interested in the topics (social ning a country-specific research agenda. dimnsions) treated in the market-, sector- and groupspecific Analysis Plans assembled in this Analysis of Ohe Priry Survey Volume (again, the prioitization of topics is a country-specific matter). The role of the Pover" 126 A country-specific research agenda might Profile is pivotal in this context. The Poverty start with the implementation and analysis of a Profile is the first of the micro-lvel Analysis Plans Priority Survey, followed later by an Integrated to be implemented because it dehneates the Survey and analyses at both the macro-meso and magnitude and the depth of povr - the key the meso-micro levels as discussed in this Vol- social problem in most countries. The starting ume. The Prrioty Surey focuses on symptoms point is the extent of poverty, ie., the number of and indicators of immediate need. It therefore poor people and households in relation to the has a strong focus on poverty and basic needs total population An immediate policy lik is because these are the key areas for immediate provided by decomposing that infornation over attention. Since the construction of income and the socio-econondc groups that make up the socexpenditure measures cannot be undertaken ety. The socioeoonomic groups are defined acthrough this strment, the main challenge Is cording to policy-relevant criteria and they repthe design of suitable proxy measures. Analyti- resent households which are relatively homogcauly the task is to use these meaures to identify enous in terms of the way structural adjustment target groups (Chapter Ten). affects them, both through the incomenerating 127 The policy focus of PS-based analysis is mechanism and through consumption. Therethus not so much the modification of employ- fore, one can anticipate that poverty Incidence meat, education or health policy, but rather to widl vary widely acmss these groups, because identify those groups in society which are par- adjustmenthas differenteffectsondiffert types ticularly disadvantaged in any one of the dimen- of income. The decomposition of poverty wil sions that constitute the standard of living and to thus provide the first insight into which the main document the extent of the disadvantage. The target groups are for poverty programs and into analysis wil thus not necessarily provide indica- the extent of leakages that may occur under diftions about the type of needed intrventions to ferent tamng options. solve the problem in the long term. This indeed 1.30 On ethical grounds, one cannot make a is the role of the sector-specific Analysis Plans. distinction between people who have been poor Here the aim is to set priorities based on obra- a long time and those who have becone poor or tions of the extent of a problem in different socio- destitute as a result of structural adjustment. economic groups. For example, if incertaingroups Therefore, the aim of the Poverty Profile is not to malnutrition is severe, this provides infomation distinguish betwetn the two, but rather to pro- 16

26 vide t relevant characteristics, that Is, a profile, countries the public sectes wage was, for a long of al poor In tthe soety. lhis Poverty Profile time, well abovemrket-dearkntenes. Inrecent thus may not be intrpret as represnti a yearstssituationhasdisappearedduetosalary causal link to adjustmne Indeed, adjustment erosion. Moreover, in many countries the public creates both wies and losera The currtly sector acts as the leading sector for wage deterpoor, as obseve by thedata, Wude the people minatton in the private sctor. The lattr frewho were poor before adpjstmt and thwse who quently exhibits inflexible wages, so that the bulk became poor theafter. However, by defiton, of adjustnmnt occus rough changes in the volthe gainers (i.e., the people who were poor before ume of employment rather than wages. AMl in all, adsbnent butwhohavesuccssfwllyresponded adjustment can be expected to induce a greateto the new incentives and are now no longer mobility of labor and to reduce job secuity. Tu poor) are not included. means that a greater number of people are bal- 131 Poverty Is not just a question of having an Ing secondary jobs and/or that labor turnows expenditure level below a cut-off point, but also has increased. The extent of this phenomenon of having an inadequate satisfaction of one's ba- and the types of workers most affected (male or sic needs: educatior, health and housing. An female, educated ornon-educated) isamajorarea important task of the Poverty Profile is therefore of analysis because it will help determine the to identify whether the poor, and in particular shape of the employment-creation programs. the destitute, are also sufferigin the fulfillment 135 A group that will deserve specal atten of those basic needs. If it is found that this is the toniistheunemployed-thosepeople whohave case, a double tleting of progms may be not managed totake successful advantage of new needed. Flrt,prograrsneedtobetargetedwhich opportunities and incentives provided in the lacan improve the productivity and the income- bor market. The question of youth employment generation potential of the poor, but second, pro- is particulry strh8entl In many countries the grams must also be more directly oriented to- output of the educational system does not meet wards immediate alleviation of health and nutri- the demands of the labor market, and the education deficiencies, lack of access to education and tional system does not respond sufficiently rapinadequate sheter. idly to the changing incentives under adtsment. 132 Beyond the development of a Poverty The extent of unemployment among recent ProfUe goverents and reserh arlikely to graduates is therefore a key indicator, and probe interested in carrying out analyses on other grms will have to be designed to retrain those soial dimensions Some of the policy issues likely people in a suitable way. to arise are sumuarzed below in the order in 136 Given the future importance of the urban whichtheresectiveanalysisplansareprented informal sectorasa creatorof jobs, the productivin this Volume. ity and profitability of the enterprises in that sec The Analysis Plan on Employment and tor deserves attention. The analysis has to relate Eanings addresses the radical chnges in the in- profitability to the characteristics of the entreprecentives for differenttypes of work brought about nerw in terms of age, education and migrant staby adjstme. In general, adjustment shifts the tus. It wil be important to monitor the sectoral incentives from sectors produdng nontradale composition of output to see to what extent engoods to those prducing tradae goods. The terprise profitability responds to e shifts in inimpact of this on rural populations will depend centives in favor of tradable goods. Significant very much upon whether food in the country in bottlenecks can occur in tenns of access to credit, questinistradableornot. Withn tradableitems, access to capital goods (which in many countries adj_usmt should promote the production of are imported and typically become more expenboth export items and import substitutes. sive under adjushnent) and access to technical 134 Another area in which adjustment affects and commercial know-how. Special programs employment Is tough estrucuing of the gav- may need to be designed as part of the adjustenunt and the para-public sector. This usually ment package to make sure that entrrneurs impies ts as prt of measures to en- have the ability to respond properly to the new hance e. The task Is thus to find sittable incentives. Institutional bottenecks can result in employment oppotunities for these laid-off a marcost when allocativefficiency cannot be people In the context of the Inentive structure acheved. put inplacebytheadjusftmentpwogram In many 1.37 Sector-specific socia dimensions analysis 17

27 is likely to focus on health, educadon and food obtain an accurate picture of the supply of health security and nutrition. Poliy and analysis issues care. However, simple associations between in these sectors are heated next. health programs and health outcomes may not 138 The Aralysis Plan on Heath suggests that measure accurately the effects of public spending structural adjustment may Influence people's on health care. The reasons for this include the health status by affectin governmental expendi- possbility that the spatial distribution of governtureson health-related servicesand market goods, ment health programs may be Influenced by enby changing the cost of these services and by virnnuntal features favorable to the occurrence altering real incomes due to changes in wage of certain debilitating diseases, which may lead eamings and farm profits. But health is multi- to positive observed correlations between health dimensional and difficult to measure. It is sug- program spending and frequency of illness. It is, gested in the Analysis Plan that anthroponwtric therefore, recommended In the Analysis Plan to measures are useful indicators of health, but estmte how changs in local programs affect drawing condusions regardig health on the ba- changes in the health of the local population sis of nutrient intake is considezed to be famught (fixed-effects estimation in order to avoid biased withambiguties.((nutrientsareipnportntiputs conclusions based on spatw heterogeneity in in the producton of health, but unsatisfactory factors determining the prvalerce of disease). ndicators of health status.) Self-reported fre The Analysis Plan on Education disftquency of illness is another proy for health. The guishes between supply and demand facors as Analysis Plan focuses on reduced-fonn demand determinants of the output of the education sysequations for health, using prices of food and temn in a country. The supply side consists of the measures of accessibity of public health pro- schools and institutions as well as the teachers grams, household assets and wages and house- and the teaching materials Structual adjusthold "public" goods (sources of drinking water, ment often implies budget cuts which may insanitary facilities) as independent variables. The dude expenditure cuts for public schoolin& inpoic,ydiscussiondevelopedintheplanisinterms creases in user charges for education, reductions of food prices and 'programs," ie., health facili- in the provision of education materials and inties, fertility control programs and household creased rationing of access to public schools. On "publice goods. the demand side, adjusmnt policies will affect 139 The availability and quality of health fa- household incomes as well as prices, which will cilities is in lae measure a consequence of pub- have an impact on the amount of education servlic expenditure choices, although private sector ices demanded. provision of health care (for example, through 1.l1 Analytically, the income effect is to be dis- NGOs) is important in many cases. It is well- tinguished from the price effect. The fonner opknown that health sector governmental spending erates through changes in income, which directly has tended to be concentrated on higher-order affect the demand for education. The price effect (ie., hospital-based, curative as opposed to pri- works through maldng schooling directly more mary and preventive) care In Africa. Higher- expensive (user fees) and/or through changes in order services tend to benefit the higher income the opportunity cost of attending school. The groups. There is, thus, a policy issue In terms of Analysis Plan focuses on the determinants of the the intrasectoral alocation of resources. Further- demands for schooling. There are three key policy more, the management of public spending in issues here: (i) How much will the ability of health, and program input combinations, are of- households to send their children to school be ten deficient, leading to low and, in many cases, affected by a reduction in their income? In other reportedly still deteriorating quality of health words, whatistheincomeelasticityof thedemand services. A question raised in the Analysis Plan for education? (ii) What is the impact from the is whether the reduced-form health demand price effect? Adjustment programs frequently equation accurately measures the effect of public include and increase user charges for education. proams on health outcomes. Official data on (iii) What is the long-term impact of adjustment progmpla mentandexpdituw onrecurrent on the demand for education? If adjustment will inputs by district and Community Survey data tmly generate growth, then there should be an on actual availabilty of health facilities and com- increased demand for education due to an inplementary inputs (drugs, disposable materials, crease in the expected retuns to education. health workers) may need to be comnbined to 1.42 If a household is faced with a reduced 18

28 demand for education, this reduction will not targeting are well-known. Self-tagetingthrough necessarily be uniform for different types of Inferior commodities (in a consmner preference, household menbers. The demand analysis must not a nutritional sense) holds considerable poat a minimum be differentiated according to gen- tential to limdt admistative cost and leakage of der. In doing so, the analysis of course overlaps benefits to non-needy consumers. There are, with that addressed in Chapter Eight below on however, few successful expeiences of self-tarwomen and structural adjustment. geted food subsidy schemes in Africa, and op- 1A3 It is important to note that the impact of portunities for self-trgeting should be explored. adjustment on education is not limdted to the 1.46 Social dimensions analysis, finally, indirect impact on the educational sector itself. cludesgroup-specificinvestigation. Theanalysis There is an important indirect impact through of women and smallholders (two groups which thelabormarket, thesmooth functioningof which partly overlap in Africa) In the context of strucdepends on a steady supply of suitably schooled tural adjustment is therefore taken up next. worker It wil also affect the ability of mall- 1A7 The Analysis Plan on Women starts off by holders and informal sector entrepneurs to in- pointin td the fact that adjustfmnt changes increase their productivity and take advantage of centives and constraints for the generation of difnew opportunities expected to accompany the keret ypes of income. It is quite conceivable adjustment pros. Lastly, education has an that the constraints faced by men and women, impact on health and nutrition status. and therefore their ability to respond to those in The educational analysis therefore pro- cenfives, are not the same. On the demand side, vides not just an input into the design of educa- men and women are not likely to have the same tion policy, but also the detenmination of policy propensities to consume goods and services such vis-a-vis nutrition, agriculture and other sectors. as education and health Therefore, when faced Although the education analysis plan does not with falling incomne and/or rising prices, the specificaily address the internal efficiency of consumption of goods and services and basic education delivery, this issue cannot be over- needsachievementsmaybedifferentiallyaffected looked. Indeed, even though Ministries of Edu- for men and women. Hence a gender distinction cation may be faced with budget cuts, the same in analysis and policy design is warranted. amount of money can be used to deliver very 1.48 The supply response of individuals, both different packages of services. A package which men and women, depends in essence on the is mainly oriented towards higher education will amount of resources they control The two key be biased against the poor. On the other hand, a resoures are labor and credit. Because housepackage which strongly supports universal pri- holdsdonotnecessarilyfunctiondemocratically, mrwy education and the provision of adequate women may not have complete control over their teachingmaterialsforprinwyeducationwiflhave time. The traditional division of labor may dica strong egalitarian impact. tate that they spend sigrdficantime on specific 1.45 Policy issues in the area of Household food functions such as fetching water and wood. security and nutrition include (i) the analysis of Consequently, ft must be known (i) what the adjustment-induced changes in wage and pro- patter of time use of women is, (ii) whether ducer income, food prices and consumption; (hi) rural womien have the necessary control to rethe Identification of food-insecure groups based allocate their time when, for example, the relative on the foregoing and an analysis of food budget prices of difterent crops change and (ioi) whether shares, expenditure elasticities and (where pos- women can gain access to the necessary inputs sible in view of data constraints) the comparison such as land and credit. of food consumption with "requirements; and 1.49 A key question is the role of food in the (ii) the various market and non-warket forms of adjustment process. If food is a tradable good, intervention available to governments to raise then major opporbtities will open up for women. the food security and nutrition of specified groups. However, this also means that resource mobility These interventions include targeted indirect in- will be highest among women, and the question come transfers in the form of commodity-specific is whether the cunrent functioning of the labor subsidies, public employment shms and di- market wid make that re-allocation possible. Inrect income transfers in the form of food aid, cash stitutional, legal or other constraints may exist or food stamps or ration cards entitling the holder which wil impede this re-allocation, resulting in to subsidized food. The practical difficulties of allocative iefficency which may need to be ad- 19

29 dressed through speal progmams In rural areas interenon or other compensatory assistance an inportant element wll be extension servi. may be required by some grops These rvices are often biased in favor of men, 153 Anoh poicy issue arises when poorer and to the extent that new technologies and pro- goups fail to benefit fom adjustn as producmodon of new crops occurs ftrough these ser- ers because price signals are not trnmtted to vices, women may have less access to them This them due to infrastructural and institutional may require tedespgn and re-targeting of exten- shortcoi or because of an nadequateiitial sion services. flow of consumer goods to the countryside. t is 1.50 There is a complex interplay between the weli4a n thatagrlculural markeft processes use of health and family planning services and are less efficient in Arica than in Asia (measured access to employment and education. It will be in tenis of the proporton of the consumer price crticalry important o look at ates of non-enrol- of food grains paid to fanmers). Thus, if smallment in prinmay schoolin& where such rates ae holders fall to shift to profitable (ie., tradable) frequently twice as high for girls than for boys, crops or to Increase the relative importance of thereby creating a disadvantage for girls which ths commoditis in their output mix, they canthey may never be able to overcome. not gain from adustment as producers Public 151 The analysis of policy issues regarding investment in agriculture and the development Smalholdes starts from the realization that under of instituions capable of servrig the needs of given agrecological conditions, the income smallholders must accompany programs of possibilities of smallholders are largely deter- macroecnomicreform. nuned by their labor use and their cropping choices, which, in turn, are influeed by relative Princple of anaygis prices in relevant markets and the availability of labor. (Land scarcity is in many African setings 1.54 As stated above, the selection of social dinot a binding constraint to production and in- mensions analysis topics is a country-specific come possibilities.) Smanluolders typicay pur- matter. It is expected that the Analysis Plans sueacombinationofactivitiesincluding off-farm presented in this Vohlume serve the purpose of wage employment and work on their plot where asstng country-based researches in their task they produce tradable or nontradable crops and of assessing the social and distributional conseanimas or, more ly, a combination thereof. quencesof strucral adutattand longer-term The relatve impornce of these activities as de- deveopment policy. The Analysis Plns discuss teninants of total income may be used to clas- bothnehiodsof descriptive,tabularanalysisand sify the smallholder population Into predomi- multivariate investigation. Both types of analynantly export crop producers, producers of trad- sis are likelytobe required in countrybased SDA able foods for Import-substitution, rural laborers reseaich, Caeul attempts to set out the tabular and -subsistence" fames, e., rural families de- analysis posities are made in particular In riving a particularly small proportion of their Chapters Four and Ten. Space limitations preincome from exhange activities through labor or cluded an exhaustive presentation of sample product markets. tables (called"fme in this Volume) in all the 152 The key policy issues related to small- Chapters Readers are encouraged to develop holders under adjustment are the evolution of their own sample cross-tabulations in their social rural wages relative to product prices, the nature dimensions field of partcular interest. Given the of changesin labor use, other inputs and the farm complexity of macro-micro linkages, some multioutput mix It is stated in the Smallholder variate analysis is indispensable in most situ- Analysis Plan that rural wage labor (which is aons. The frequently inevitable omission of imexpected to benefit from adjustmnt in the longer portant expblnatoy variables from crosswtables run) might suffer a real wage fall in the short run and the fact that interaons can be studied only This could occur if larger famers producing for rudimentarily through tabular analysis can lead export or domestic market consumption were to misleading policy conclusions in the absence slow to expand their production. It is also not of regression technques. An attempt was made certain whether smawolders would require a in this Volume to limit mathematical expositions real wage fal to induce their epansion of tradable to a minimum, to present them in speci boxes production. The short-term effects of devaluation so as not to interrupt the flow of the text and to on real wages are therefore ambiguous, and food provide verbal explanations for all points of 20

30 modeluing expressed in the form of equatons. might, th^ezdo, find It usefu to glean comple- Wile e various uhpters are desped to be mentaryanalytcalguidacfrmchaptsothfer self ndlm stofthmaplyareducd-fornm than the one(s* in which he or she is primarily app the estmaton of outcomesasa fnc- Intersted. tion of household-cholce variables. Te reader l~~~~~~2 / 2

31 Analyzing the meso-economic effects of structural adjustment Introduction macro-meso analysis. This is followed by a discussion (in paragraphs 2.23 through 2.28) of the 2.01 Tis Chapter is concemed with the real- dsinctionbetweenexportables,iznportables and economy effects of adjustment policy, and how nontradables - a distinction which is central to these might be analyzed in the Africn settg. any assesment of the effects of adjustment Therearetwomainreasonsforthlsanalysis: First, 2.03 This Chapter differs in one important remost governments need to understand how pol- spect from the others of this volume. It relies icy reforms affect their economdes, and elements generaly on data sources that are extemal to the of the SDA information systev, notably the comn- SDA household surveys, on which the other munity survey, can enhance the data base for chapters are based. It is therefore difficult to set such an analysis. Second, as part of the SDA out an analysis plan with the samne precision, analysis package, this chapter serves as an essen- since what is done in each of the partcipating tial input into the other Analysis Plans descibed countries wil depend critically on the available in this Volume. Our main emphasis wil be on data. Consequently, the chapter suggests two the latter objective - the chapter wfll seek to broad types of approach, each having quite difexplainhowsomeoftheeconomicvariablestaken ferent data implications. The first approach (deas explanatory variables in the other chapters of scnbed in paragraphs 2.29 through 2.67) assumes this Volume are themselves influenced by policy ordy a fairly basic data set -essentially the nainterventions. The sub-set of explanatory van- tional accounts, public sector accounts and some ables with which we shall be concermed are at the price data. Paragaphs 2.29 through 2.35 review mo-cnomnic level. Thesearethe variables which the process of adjustment. It applies the distinccome between the maro-policy interventions, on faon between expenditure-switching and expenthe one hand, and the micro-level (or more cor- diture-reducing effects, and introduces a simple recly in this context, the household-level) vari- method for assessing the weight given to each in ables on the other. the adjustment program under, consideration The next section places the macro-meso Paragraphs 2.36 through 2.60 review ways in analysis descrbed in this Chapter into the con- which expenditure-switching effects can be anatext of the SDA analysis program as a whole lyzed, whilst expenditure reduction is treated in which is the conctrn of this Volume. A brief paragraphs2.61 through A second approach, sumnmary of the ways in which adjstment policy involving modeuing techniques, whir are highly influences markelt and infrastructure is given in data intensive, is en discussed in paragraphs paagraphs 2.12 through Paragraphs ftrough Finally, paragraphs 2.85 through 2.22 then outline the essential country through 2.87make someconcludiigobservations. background information that hs required for 23

32 The context of maro-meso analysis by specfying a simple two-stage analysis deslg In the first, the analysis is coerned with how 2.04 Much of the prewious work on the social policy Iteventons have affected markets and dimensions of adjusttent has failed to providea If t -which we ihall term nacro-meso convincing analysis of the effects of stuchu analysis. This is the cner of the present chapadjustment. Thdsisbecauseofageneralfailureto ter. The second stage deals with meso-micro separate out the effects of the adsusent policies interactions, and these are the coneem of the reper se from other factors operating on house- maning chapters of this Volume. The interacholds concuretly, inluding the effeds of the tions between these two analysis sges are illusshodcs to which the nacro-policies are adustig. trated in Figure 2.1. Many studies have been explicitly agnostic about 2.05 At the top of the figure, five broad categowhether the deterioration in the welfare indica- ries of 'mao-policy' interventions are lsted. tors it traced were due to the recession itself or to Some of these are macro interventions in the pure the adjustment policies that followed (see, for sense of that word - such as monetary polcy and exanple, Comia et al, 1987). SDA policy analy- exchange-rate adjstments. Ohers are applied sis, however, should go sonw way towards at- at the sectoral levd - including institutional retrbutingeausationinitstreatmentofadjustent- forms and trade policy. The policy invenhuman dimensions links. To achieve this, the tions are combned to achieve major macro-polproposed analytical methods treat explidtly the icy targes, typically ncreased economic growth, conduits at the meso level which trmit the and internalandextemalbalance(correctinghigh effects of adjustment policies to the households, inflation and utnstainable budget and external these being markets (both factor and product, of- deficits). They are also appled to improve effificial and paralle), and social and economwc in- iency (through resource re-allocations) and eqfastrudture (see World Bank, 1990). This is done uity (though poverty alleviation). These policy Figure 2.1: Macro-meso-mlcro analysis MarkIt Infdasi ucture Lab"r Product Credit Soda EcOnomC 24lousehoU Hueltdd Housdd l inexnre a5sptf exedaths Demand 24

33 Ins ents are shown influendng the two key mawketing fhilties, ad so on). b ( = 12,3) are elrments of the meso-economy - markets and the parameters to be estimated. The meso-level lrastructure. Three main market effects are variables as used in the meso-micro analysis plans noted in Figure 2.1: labor, product and credit are obtained from both the Commtuity Survey markets. The figurealsodistinguishestheefcts (CS) and the Integrated Survey (IS) of the SDA of nacro-policy on socal and economic infra- infomtion systm In someases (for example, structure. health and education series, and maket pries), 2.06 The sigrdficance of these meso-economic data are obtained dirly through the CS. In effects for household welfare Is also illustrated in others, the information is obtained from house- Figure 2.1. This defines (in thebottom box) some hold n nibers, through the IS questionnir Such observed outcome at the household level which questions as "Have you had any difficulty in in some way reflects household welfare. This obtaining credit over the past 12 months", can might be income, expenditure, poverty, educa- provide some evidence (albeit through the eyes tion, health, status of women, and so on. These of the household member) of meso-level chsracoutcomes are shown to be the result of the com- teristics, in this case the presence of rationing in bined influence of supply and demand factors. the credit markel Given the once-off nature of Thus, for example, education status of household these surveys, crss-section data must form the members (an outcome variable) wiibe influenced basis of estimating equation (1). by the availability of schooling (a supply factor) 208 Thethirdstagesekstoassesshowchanges and the income of the household (a demand fac- in the X 2 and XY variables arising from adjustment tor). Notice that markets Influence both supply affect the dependent variable, and thereby, and demand sides of this equation. Similarly, household welfare. This is given by, infrastructure will affect both supply and demands. Thus, for example, changes in the labor AZ = b 2 (AX2) + b 3 (W) (2) narket will affect the numbers of teachers available, and influence the supply of education ser- The changes in the meso-level variables induced vices. Changes in product markets may influence by adjustment (AX, i 8 2,3) will themselves inhousehold income, and thereby the demand for ducechangesinthedependentvariable (AZ). This education. Social infrastructural expenditures by stage is certain to be the most difficult and chalgovenments will have a direct effect on educa- lengingundertakingof thesdaanalysisscheme. tonal supply factors, but they may also affect Theraindifficultyliesinquantifyinghowmesohousehold incomes, and thereby influence edu- level vanables have changed asa result of adjustcational status through demand for education. ment policy interventions. This is the chalnge 2.07 Typically, the meso-micro analysis plans of the present chapter. It seeks to identify the follow three nain stages. First, they define the main meso-level variables used as explanatory dependent variable of interest. This obviously variables in the analysis plans presented through depends on the subject at hand, and on the par- this volume, and to suggest how induced changes ticular dinsion of household welfare of inter- in these variables can be estimated. est (for example, smallholder income, education, 2.09 The network of influences illustrated in health, etc). Second, a series of explanatory or Figure 2.1 is the context into which the macronght-hand-side variables are defined, and these meso analysis plan in this chapter is to be placed. are used to explain variations In the dependent An important implication of this for the present variable. This relationship tkes the following chapter is that we are not necessarily interested general form: in all the effectsof adstment policy. SDA macromeso analysis must focus on those elements of Z _ a + brx, e b 2 X 2 + b 3 X (1) the meso economy which are considered of particular significance in determining household where Z is the dependent variable of interest, X, welfare. Inotherwords, the analysis is not meant is a vector of predetermined household assets in any sense to be a fully comprehensive account and characteristics, X2 a vector of market vari- of the way adjustment affects real economies. It ables, such as prices and dummies indicating is essentially selective, concentrating on the key quantity rationing, and X3 a vector of infrastruc- right-hand side variables of the meso-micro tural variables (such as dummies for health and analysis plans. education services, communications, access to 2.10 Figure 2.1 Ilustrates only the broad prin- 25

34 Table 2.L- Mso4level variables utilwd in h_psehold data aalyss meao-iro analydi - -SWlsEuiye h Hsfh u Marketvarabhles Pmduct na*et prices '4 '4 44 inutprles I 0 * Market rationi * 4 Market wages *4 * ' 4 N Wag Ifexibiultles Intersates * 4 Access to aeit * 4 vinastucturl variables Roe.scommu,clns *- * Irrhto services ' *4 MarketingnsotItutins 4' * '* HeGefii servies ' 4 Edai services Othersodalservices * * i dipes which govern nmacro-mo analysis in the A review of the meso effect of adjustment context of the overall SDA analysis program It does not identify which meso-level variables are 2.12 A progam of poicy reform usualy inof particular concern to the various meso-micro volveschanginga wide selection of policy instruanaysis plans outined in thds volume. Table 2.1 ments, raging from the purely macreomic attempts to bring al this together in a simple (such as monetay and exchange rate policy), to analysis matrix, by identifying which meso-level sectoral policies (trade and priding polices), and variabes are utilized in each of the analysis plans. even to policies at the micro level (as for eample, A double star in the relevant cell indicates that in the case of re-trining schemes for retrnced the variable is considered to be itical for meso- public-sector workers). A compaion volume micro analysis. A single star indicates that the (World Bank, 1990) sets out a ftamework for anavaiable is utlized In the appropriate chapter, lyzing how chges in these instruments (even but it is considered desirable rather han critical wacro-instrumentsuch as money supply) are The wide range of subjects examined by meso- liely to influence the real economy. Our purmicro analyses imphes that a variety of meso- pose is not to repeat these argments here, but level variables is caled for. However, the double simply to rehearse the main condlusions, which star ratingggests that a more,limited selection is sufficient preparation for this macro-meso of key meso variables can be Identified. These are analysis plan. product prices, wage rates, credit, agrcultural 2.13 Our review of the meso-micro analysis services, education services and health services. plans established two important elements of the These are the main meso variables considered in meso-economy-makdetsandinrhastucture. In thischapter. what ways might adjustmen policies be expected 2.11 Before we embark on analyzing how ad- to influence these two elements (which are in susment might influence these key variables, turn expected to have some significance for some preliminary analysis must be onducted on household welhfre)? Beginning with makets, tree iportant issues. Te first conwerns the we qan define the following major effects: efects that adjustnt policy can be expect to * Most adjustment polides attempt to encouhave on meso-economic outcomes, in principle at age resource trsfers out of nontradables and least. The second involves essential country into'tadables. To achieve this, the real exchange background information, and the third deals with rate (defined as the price of tradables relative to the operational distincton between the various nontradables,orpv/p)mustbeincreased (known classes of conmodities used later in the chapter. as a real exchange rate depreciation), either through a nomi currency depredation (which 26

35 raises P), or through domestic monetary contrac- tion can have direct effects on wage detmeition (thus reducing P). If a govemnment is suc- tion. cessful in changing this fundamental relative 2.14 In these ways, an adjustuent progam can pie, i will give strong signals for resource re- have profound effects on relative product and allocation. These signals will influence factor factor prices. Macro-meso analysis must, above makets, the real wage falling (ising) if tradables all, be able to offer some insight into how an are relatively less (more) labor intensive than adjustment program in the country in question nontradables. In the short rn, the real wage has affected the main markets and relative prices chang will depend on workers' consumption of the country. Clearly, this cannot be achieved propensities. for every specific factor and every product - * In addition to operating on P,/P,, a gover- theremustbesome aggregation. There arestrong ment may also change relative prices within each analytical and empirical grounds for aggregating of these product categries, and especially within products according to their 'tradability' in world tradables. For example, trade liberalization can markets. Thus, nucro-mo analysis must at least be expectd to reduce the price of previously distinguish between exportables, proteted improtected importables relative to exportables and portables, other importables and nontradables. other importables. Again, these relative price The issues raised in making these agegations in signals will induce resource transfers, and will practice are dlk.'ussed in paragraphs 2.23 through lead to changes in factor markets. For example, 2.28below. real wage changes can be expected, depending 2.15 The meso-micro analysis plans presented on relative factor intensities in the sectdrs. in this volume also find infrastructural variables * An adjustment program can include any to be important in determining household welnumber of specific price policies (involving fare. And so it is important for macro-meso changes in consumer taxes and subsidies, and in analysis to establish how adjustment might inthe polcies of state marketing boards) which will fluence the provision of infrastructure. Most furthermodifytherelativepricechanges induced (though not all) infrastructural services are prothrough the real exchange rate depration and vided by the state. This means that such services trade ieraliation. are likely to be cut as adjustment seeks to transfer * Policy reforms will not only influence the resources out of the nontradables sector, an imstructure of relative prices and resource alocation, portant section of which Is the government itself. ihey may also influence resource utilization. Moreover, the need for fiscal discipline during Whether adjstment leads to an increase or a adjustment often reqires a reduction in governdecrease in resource utilization depends upon ment expenditure, which also may conflict with which of a number of opposing influences is the the provision of infrstr services. stoger. On the one hand, there are factors which 2.16 For these reasons, a period of adjustment shoud improve resource utilization - the in- may witness sigrnfict changes in the provision creased availability of foreign exchange, remov- of economic and social infrastructure by the state. ing rationing in both the foreign exchange and The net effect of the provision of these servics goods markets, increasing real wage flexibility, depends on whether the private sector 'crowds and enhanced investment opportunities and ex- in' - that is, steps in to provide the services pectations. Set against these is the need for aus- through the market. For example, if the state terity during adjstent, which generaly -im- provision of education and health care is cut, are plies that govemments must reduce fiscal defi- there compenseting increases in private schools cits. Such policies generally reduce resource and health clinics? Macro-meso analysis must, utilization. thefore, attempt to establish the extent and in- * In addition to these direct product-market cidence of expenditure cuts, and to evaluate the interventions, adjustent can influence the oper- implications for the provision of social and ecoating rules of the market through institutional nomic infrastructural servi chages Privatization (especially of marketing Essental background informton Isitutions) can lead to significant changes in market prices. The removal of controls in credit 2.17 There are two main reasons why each markets can also cause an increase in interest countuy nacro-meso analysis plan must begin rates and a change in the whole structure of rates with some background information of the type in the market Sidlary,changes in labor legisla- set out in this section. FiAst, the darteristcs of 27

36 counties particpating in the SDA Pgm are categolies, if posible identfying consu -, incertain to differ - in tms of their populations, tennediate and capital-gods imports. Regulaeonoic stucture, trade paftens, istutions tions on foreign excne nmight also be identiand economic ystem To undetake a meng- fled, along with a review of the exchange rate ful macro-meso analysi requires some prelimi- system in operation. nary undrtanding of these underlying country 2.20 Finaly, background data are required on charactic Second, each country will have the institutionalc attics of the country, covfaced different economic Imbalances and to dif- ering mainy its economic institutions. These ferent degrees, and wil have responded in dif- include the nature of product and factor markets, ferent ways to corre the imbalances. Again, to the size (in terms of employment) and activities properly analyze the effects of adjustent on the of the formal and infonnal sectors, the incidence key meso variables requies some prepartory of government controls and Interventions in work, establishing the periods of destabiisatlon markets, the extent of public ownerhip, the and adjustment, and indicating the principal fa- 'depth' of the financial system, and so on. Backtors underlying both, are necessties. grund dat should also be prnted in the soial and econic infrastructure of the country Country chactristics -describing the main f oms of conmunications, recent trends in transportation, extent of electrifi Information on the main economic and cation, Irrigation services, and the main institusocial dharacteristics of each country should be tional arrangements in the provision of health compiled as an important preiminary to macro- and education services. meso analysis. Some of this information may be available in the Country Poverty Assessment Periodsf destabifiton and adjushnent Paper, prepared under the SDA initiative for the country, but we repeat some of the more impor Information is also required about the petant pieces of information which are required riods of destabilization and adustment in the here. The first set of tables which is needed deals country concemed. What is done here is essenwith economi grwth and struure over the past tialy country specific, and depends on the tim years or o Estiomates of GDP, and the ing of the SDA policy initiative in the country in gsrowth of GDP, population and per capita GDP relation to its recent economic history. f the data should be presented. This should be acodmpa- collection and analysis work occur after an adniedbyabrief accauntofstruchtal changeinthe justment program has been in plac. for some c-oy, if possible showingchangesinthestruc- time, it will be possible to identify periods of te of production and employment over the same destablization and adjustnt The period of period (or preferably, for selected years over the destabilzation is one in which major unsustainperiod). The following disaggregation should be able macreconomic imbalances emee, creatsufficient: Ing the need for the adjustment program which primary production (of which for export) followed. During the adjustment phase, one * mining would expect to find these imbalances being re- * manufacturing duced. Thus, during periods of destabilization, consumer goods inflation rates, external deficits and/or governintemuediates ment budget deficit should incease. Tere might caital goods be other symptoms of the imbalanc, such as an d consruction acute scarcty of foreign echange, the imposi- * udlities tion of import restrictions, a shortage of imports * serices (of wichh govnment) (of consumer, intermediate and capital goods), 2.19 A third category of background data con- anappreciation of the real exchnge rate, resource cerns the tde and ehange patterns of the coun- transfers into nontradables, and so on. These tby. Tables should illustrate the nain changes in should all be eamnd In order to establish whh the exteral accounts over time, showing the petiod is best considered as one of destabilizapricpal sources of foreign exhange on current tion. It mdght also be helpful if some account of and capital account. This dsould identify the the causes of these imbalanc is given. This mainexposandsourcesoffinance. Informtion would be possible if the analyst can draw on should also be presented on the-main import existingworkinthisar. Itisnot recommended 28

37 that original research be conducted into this is- be eogenus. One of the more important probsue under SDA mao-meso analysis. lems encountered in using this type of dassifica Similarly, the analyst must idenify the tion of product markets is that ommodities can period(s) of adjustment, and the policy instru- switch categories, frequently in response to tlhe ments which were (or are) used to achieve the policy change under invesfigation. The most adjustment. This will involve a brief historical important reasons why goods are nontraded are account of the main policy reforms enacted, in- c ial policy (eg. prowhibition of imports) duding major fiscal and monetary contractions, and transportation costs. Taking the transportaexchange rate adjustments, price and marketing tion mark-up to be q, and the world price of a reforms, and institutional changes (such as prl- commodity to be P*, the domestic price must be vatizations). It is important to establish both the equal to or less than P*/(l+q) in order for it to be timing of the application of the policy changes exportable (assuming no trade taxes/subsidies). and their effects on the major macro-balances - Simiarly, for the commodity to be importable, its the trade and budget deficits (usually expressed domestic prce must be greater than P*(1.q), as as a percentage of GDP), the rate of inflation and otherwise its importer would not be able to comthe rate of economic growth. The penods of pete with domestic supplies Thus we have a destabilization and adjustment may not be as range of domestic prices for which the commodclear-cut as the above discussion would suggest. ity is nontradable - neither an exportable nor There nay be elements of policy reform during importable. This range is simply given by, periods of destabilization, whitst some measures of macro performance (for example the rate of PP/(l+q) <P. < MI +q)sp 1 (3) nflat) may continue to worsen during periods of adjshnent (when other indicators show an 2.25 The difficulty here is that a comumodity improvement). The analyst must use his best can cross these boundaries and move from being Judgement in such cases. But it needs to be em- nontradable to being either an exportable (if the phasised here that the macro-meso analysis pre- domestic price falls sufficiently) oranimportable sented below cannot properly be instigated (if the price rises). Country-based empirical work without some broad historical assessment of this will clearly have to identify where such changes sort occur. A second difficulty is that commodity classifications may change geographically. A Exportables, importables and nontradables certain conmodity may be importable at or near the port of entry, but as transportation costs in The distinction between exportables, im- crease its price in remoter areas, it may become portables and nontradables is central to the over- entirely insulated from world markets. This may au eonceptalization of the SDA policy initiative, have become more common during periods of since most adjustment policies can be expected to destabilization and adjustment, since transportainfluence each product category differently (see tion networks in Sub-Saharm Africa have dete- World Bank 1990). One of the essential founda- riorated, and costshaverisenacmordinuy. Again, tions of the analysis program of the SDA, the- the SDA Prgran will be required to mnake a fore, is to operationalize the distinction at each carefulassessmentofhowall thisnayaffectpnces country level. Whist this is essentially country facing households, especially if such groups are specific, some general guidelines are presented located in remoter areas not well served by here. physical infrastructure Nontradables are those goods and serv There are sectors whose outputs dearly ices whose prices are determined by domestic fall under the 'tradable label, such as producsupply and demand. This is due to the nature of tion of cash crops for export. Similarly, many the good involved (eg. public services, housing government services are unquestionably nonand constuction) or because transport costs pro- tradable. In between these pure cases, there lies a hibit either the import or the export of the good grey area of conceptual ambiguity. Any sector in questn, and insulate it from world markets whichisprotectedby severe inport quotas shoud Trdable goods are those which cross frontiers beindudedunderthenontradablecategory,since and, in theory, their prices are determined di- changes in the world price will leave domestic recdy by world market conditions, so that for a prices unaffected, and will only influence the "small" economy, tradable prices can be taken to margins obtained by unportes In Chapter Eight, 29

38 below, it is suggested that a useful startng point (as illustrated in Figure 2.1), they are ikely to be for most African countries would be to consder particularly vulweble to adjustment polices manufacturing as protected importable (assun- whkh emphasize expenditure reductions. A ing that the good in question is not subject to contractionary monetay and fiscal stance is lely highly restrictive import controls), and export- to lead to cuts in government expenditure, which agriculture as unprotected exportable. With the may reduce resources available for econoric and nontraded sector comprising mainly construc- social infrastructure. tion, services and the public sector, the main area 2.30 As a first step towards understanding how of ambiguity lies in classifying food agriculture. relative prices and infrastructure night be influ- How food production is classified will need care- enced by the adjustnent effort, some attempt ful assessment for each country case. must be made to assess whether adjustrent was 2.27 Macro-meso analysis must establish for achieved mainly through switching or through periods of destabilization and adjustment, *tis expenditure reduction. This can be achieved categorization of commodities. This must in- through a simple decomposition exercise. But volve the identification of any changes over time first, It is necessary to outline some underlying in the classification. For example, during the theory. early years, possibly during periods of destabili An analysis of the effects of adustment zation, food might be a nontraded good, simply polixy must respect some basic macroeconomic because of the import protection food producers identities. These are the National bncome and received. But as poiicy reforms are applied (pos- Product Accounts (NIPA), which measure the sibly a reduction in food import controls com- flow of goods and services and incomes in the bined with a devaluation), food might switch to economy, the Balance of Payments (BP), which beodme a tradable. One of the objectives of this measures the flow of current transactions with analysis is to Identify such switches over time, the rest of the world, and the Monetary Survey 2.28 The analysis must also identify any geo- (MS)', which measures the flow of money creagraphical switches in these categories, the evi- tion. The key point to be observed here is that dence coming mainly from the SDA CoMMunity these three sets of accounts are interdependent - Survey. Given information on prices at or near any one can be derived from the other two. For the ports of entry, and estimates of transporta- simplicity, assume that all borrowing and lendtion mark-ups, it should be possible to predict ing abroad is undertaken by the goverment. what prices across the regions should be. Major The NIPA identity is given as, departures from the predicted prices would suggest that the commnodity in question is a non- C +I+(G +G;)+(X-M)BC+S+T (4) tradable. where C is consumer expenditure, I is gross pri- The process of adjustment vate domestic investment, G, is government consumption (j = c) and investment (j i) expendi The two main meso-economic variables ture, X and M are respectively the exports and which are utilized in the meso-micro analysis imports of goods artd senrices, S is gross private plans considered in paragraphs 2.04 through 2.11, saving and T is total tax rvenue. GNP measured are prices (for products and factors) and infra- as aggregate product is given on the left hand structure. How are these influenced by adjust- side of (4), whilst it is measued as aggregate ment policy? The extent to which relative prices income on the right hand side. Re-arrangig change as a result of adjustment depends on the gives, degree to which expenditure swithg characterizes the policyrefomn that are implemented. If S-I Iw(G., Q + - (X-M (5) the adjustment program involves a sigwificant exchange rate adjustment, and changes in price The left hand side of (5) gives the set flow of and trade-policies, relative prices wilt change, saving from the private sector into the financial and act as signals for movements in resource re- markets and the right hand side sums the allocation (generally to exportables and unp- govermentsdemand for deficit finance and the tectedimportables,andawayfromnontradables). foreig sectdor's demand for funds to finance its On the other hand, although infrastructural serv- cunt-account defcit. ices may be influenced by the relative price shifts 2.32 The BP identity is given by, 30

39 X-M+ALd 8AR (6) adjustmenlt lhese calculations are very staihtforward, and yet they provide considerable inwheme denotes government liabilities abroad, sight into the proses of the adjustment pro- Rlsthelevelofforelgnexchangerese rvs,andas ga. First, the data should be compied along before, A denotes changes in the variable indi- the lines of Frm 2.1, which decomposes the cated. Here the cumrent account surplus plus change in the trade balance into the changes that govenmment borrowing abroad sum to the chnge occur in output (AY) and the various expnitue in foreign exchauge resrves at the central bank. components as given In (9). The sum of the 233 Fnally, the MS identity is written as, changes In output minus the changes in expenditure will just equal the change in the trade deficit. AB a AR + AL$. (7) These changes in the values of each of these items should also be expressed in percentage terms - where B is the monetary base and L the liabili- the changes in output and expenditures being ties of the goverment to the centwaank. No- exprsd as a percentage of the total change in tke, the MS identity can be derived from the the trade balance. Notie alssthat the table sugother two identties. Net saving by the private gests a memorandum item, in which the changes sector, the left hand side of equetion (5), can go to in the trade balane are aggrated over years in additional holdings of noney balances (B) or to order to capture key peiods. Thus, for example, holdings of govenunent debt (L. Similarly, the the memorandum item in Fame 2.1 decomnposes govemmentdefict((gc+gs)-t)canbefinancedby the change in the trade deficit over the periods sales of debt to the private sector, the central and , assuming that the first pebank (L"), or abroad (L,,). Thus for each item in riwd is one where the trade deficit worsened the NIPA identity we can derive a correspording (hence the negative sign) - this being a period of cne in assets and liabilities held by the institu- destabiization- and that the deficit narrowed tions involved. Thus, equation (2) can be rewrit- in the second (adjustment) period. ten, 2.35 The purpose of Frame 2.1 is to show whether the changes in the trade deficit (and (AB +ALW = (ALw +ALv +Ag + (AR- AL) (8) esecialy the Improvement in the deficit durig periods of adjustment) were associated with a It should be obvious also that (8) is equivalent in change in output (which would be an expansion an account sense to the MS identity (7). These in output during the adjustment phase), or accounting relationships form an excellent basis changes in aggpate expenditure (that is, defor comparative work on adjustment policies clines of expenditurmes during the adjustment peacross countries, since it ensures internal consis- riod). If the former, the adjustment is likely to tency in the macro-accounts. have involved switching policies, which raised 234 Using this accounting framework, and the production of tradable goods. But if the adgiven the availability of the national accounts data, it is possible to decompose changes in the balance of trade deficit into the various expendi- Frame 2.1 Changes in the trade balance and ture elements. From the national income and Its related identities product accounts (equation 4), AX-MJ AY JE A(X-M) Y -AC-A -AG -AG 1 (9) A (Pvt Aino (PmO Amo Mmw* _ AY-AE 'ArE 1979 (100) where Y is GNP (= C+S+T) and E is aggregate 198 (100) expenditure (= C+I+G.+G). Such calculations 1981 (10)) should be prepared in order to establish the ex- *-- 6M tent to which the adjustnent in the extemal ac (ID) counts was bmught about through increases in output (AY) or through deeases in expenditure. Memoandum iem: And If the latter, it should show which of the (oo0) components of expenditure distinguished in the (100) national accounts have bome the brunt of the (IOD) 31

40 justment was brought about nmaily by cuts in instruments change prices within tradables, then expenditure, either switching polides were not furdwerelative price caculations wll be called applied, or they were mainly ineffctve. This for (and these are dicussed below). preliminary assssment should then lead on to a 239 The are broadly two metods for commore in-depth analysis of the application and puting the real excne rate: the first involves effects of switching policies and expenditure re- taking price indices of nontradables and tmdables ducing policies. The types of analysis that are directy; thesecond,indirectmethoddeflatesthe required for each are addressed in the following nominal effective exchange rate with appropriate sections. price deflators Each is dicussed in tumn 240 Whether the direct method can be used to Expenditure-switching effects compute the real exchange rate dearly depends on the availability of the required price data The analysis of the switching effect of Prices of nontradables nay or may not be avaiadjustment must be divided into three broad sec- able in the published sources. If not, they may be tions. Fis, it nmust examine thie effects of adjust- proxied by taking either the sevices component ment on relative product prices. Second, some of the consumer price index (or CDP defltor), or assessent mustbemadeof the real-wageeffects a specific index which is expected to move in of policy interventions. Finally, some analysis sympathy with nontradable prices. The latter should be conducted of changes in the credit might inwudea prkelndexofasingle sector (like market induced by adjustment policy, construction or housing), or alternatively, a wage 237 A difficulty which is cdmmon to much of index (assuning that labor is a uajor component the analyses in this section arises from the pres- of nontradable production costs). ence of parauel markets in developing countries. 241 Choosing an appropriate index for export- These exist in product, credit and even labor ables and importables also depends on data markets, and also generally characterize the for- avadability. Providing there are no quantitative eign exchange makets. Analysts should ideally restrictions to ihtemational trade, urdt value intrace induced changes In both official and paral- dices may be used to Indicate the movement of tel prices, since these may well differ, depending tradables pres. Thus, for example, the importon the nature of the policy reforms adopted. With ables price index can be estimated using the unit some households trading mainly in official mar- value of imports (or a major component of imkets, and others In parallel markets, this dstinc- ports such as manufactus. Similarly, the extion is important for SDA meso-mico analysis. port price might be estimated using export unit values. On the other hand, there may be direct Induced chnges in relative prductpr information on tradables prices from official sources (induding the price data published by 238 One of the crucial ways in which macro- marketingboards). Suchinfomationisprefered economic adjustment influences the structure of when quantitative resictions mae unit value relative prices is through its effect on the rel ex- indices of little relevance in predictfiw the dochange rate, defined as the relative price of mestc price. On the other hand, offidal data tradables to nontradables(pt/p.). Changes in this tend to give a misleading impression of market relative price are frequently the main dring force prices since they usually omit parallel market of the adjustment effort, and are brought about prices, which are invariably significantly higher mainly through currency devaluations. The cal- than the official market price. culation of the real exchange rate is particularly 2.42 Frame22illusratethetypeoftablewhich meaningful when other tade policies (such as should be prepared, reporting the direct calculaimport controls, tariffs and expor taxes) are not tlons of the real exchange rate. The memoranchanged during the adjusnment This means that dum items, reportng the mean index during the relative prices within the tradables category (for sub-period, shoudd prove useful in gauging the example,theexportable/importablerve price, extent to which the relative price movement Px/Pm) will remain unchanged, so that we may changed as a result of the implementation of the use the tradable composite. 2 Under such cdr- adjustnentpogram. Thatis,anestimateofA(P,/ cumstanes, the main relative price change inx- P) for use in equation 2 above could simply be duced by adjustment should be captured by the difference in mean values of the pice index changes in the real exchange rate. If other policy for periods of destabization and adjustment. 32

41 Frae 22: Relatlve pdie indices of tadables 2A4 In clulati (10), the analyst wil have to and nontradables resolve three main diffkulties: the selecton of ns80 = 100) the best emnpii cunparts to P* and P, with whkh to deflat the nonal e ne rate e; the Sukuse ;as w possibility that e wiu vay across tradable com- PA VaS VAIW modities; and fte effects that tariffs, import cono a) 0) a#w 0a) trols and expot taxes/subsidies wi11 have on eave tradable pries 19 2A5 lhe first issue then coers the choice of 1980 t1o OG IO price iics to use in deflafi the nominal effective exchne rate, the latter being deined as a trade weighted index of the nominal exchange 1990 rates between the donmesfic curency and the curmanomwumttm ~rencles of each tding prtn. Three deflators 1979 um lhave been sggested and used In the empirical IN-90 Uliterature. 3 First, the consumer price index (CPI) of the country in riation to its competitors is --"Vdy a--nduat pfiemmybe used. Equation 10 then becmes, a. hlttielyanin4acfcb _ ptisumay bs e = e(cpi")/cpi' 2.43 The indixect metod of estimating changes ip,/p.over imeinvolvesdeflatingthe nomnal where the superscripts refer to world (w) and rate of exchange In the simple dutee sector de- domestic (d) values. The world index can either pendent economy world, thure are dutee rlated be te CPI of a prinpal trading parner, In which sets -of rlative prc. These are given by, case the calctation is on a b}iawel basis, or it may be a weighted aveae of a number of trading P /P = P jep(a+k) partner Cs in which case a mul tiaerd real P.P. = ep (l+s)/p exchnge rate is computed. We hall return P./P. = [P.*(1+s)/(P(*(l+k)' presently to the calculations involved incomputing the multilateral version of a. where e is the nominal exchne rate, k is the 2.46 A second approah is to replace the CPI tff rate imposed on imports, sis the subsldy/ with wholesale price indies (WPI). WPIs comtax rate on exorts, and the aterisk refes to prise mainly tradable goods. Tis has its lin-itaworld prioes. Assuamg (Initially) that s - k = 0, tions, however, since the denominator in (10) and that P./P. remains unchaged, then ex- should Idey reflect domestic prices of nonportables and importables can be combned into tradables. Even as an indicator of intemational a composite tradable commodity. It follows then competitiveness, the use of WPI is inappropriate, that the ral excne rate, P,/P is given by, sne intemational variations in WPIs tend to be small. A third approach is to deflate using GDP e = ep,*/p. (10) deflators, that is a weighted average of the GDP deflator of the countrys trading patrs divided Thisindexghoudreflecttwoimportanteffectsof by the country's own GDP deflator. One of the adjustment. FPst, it should show the direction of limitations of using the GDP deflator in computinduced resource allocations. A iseine, ora rea ing the real exchange rate is that it contains both exchange rate depreciation, shoul signal an in- tradable and nontradable commodities, and crease in profitabty in badable and therefore, trfore does not exactly correspond to equaa tendency for resources to flow into the sector, tion 10. Moreover, t is only made available anand out of nontradables. Secondly, ft reflects the nually, usually after some time lag. degree of the counfty's intemational competi Analysts will in ar probability, calculate a tiveness. A dedline in e reflects an increase in the number of estimates of e, using various deflators, real cost of producing tradables, and therore a and using blateral and multilateral eal exchange dedlineincomptivene. (Simlady,an incease rates. Following Edwards (1988 and 1989) and in eshould signal an Increase ina country's inte- Harbeer (1986), we suggest that e is best estinationl competitivneness.) mated using the WPI of a countrys competitor(s) 33

42 in the numerator and the CPI of the country concwrted as the denominator. The former gives gap between the offical and parallel rate, P.M.l Is likely to rise as a result of a devaluation. This greater weight to te price of tradables, whereas means that resource flows into exportables wil nontradables usually figure prominently in the be greater (and into imprbles less) than that calculation of the latter. In this way, the estimte predicted by the calulated value of e. If andysts of E dosdy corrsponds to equation IQ Altema- feel that such cages are significant, and likely tively, analysts nuy prefer to use only conpo- to invalidate the use of e, they may calculate secnents of the price indices - taking the tradables tor-speific values of the rel exchnge rate, using components of the price indices of tradig part- the exchange rate that applies to the sector conrters (be these CPI, WPI or GD? deflators), and cerred. This would also provide an opportumity thenontradablescomponentof the countys own to take into account sector-specific trade interprice index (CPI, WPI or GDP deflator). In this ventions, to which we now turn. way, the world price of tradables is indexed in 2.50 The third set of complications that arises the numerator and the domestic price of non- in calculating the relative-price effect of adjusttradables is indexed in the denominator. ment conerns the presence of, and changes in 2.48 This gives the 'best estimates of as, other trade interventions, such as tariffs, import controls and export taxes/subsidies. If these are eb= e(wpt /CPld important, and especially if they are changed under the adjustment program, it is essential that xandthe niacro-meso analysispromeeds beyond e, since e.= ( e WPI-,)/CPl Z"1 a~ ; WPIW 1 )/CPIrelative this index will give a misleading impression of price changes, and therefore resource flows within tracables. For example, cuts in exwhere the b and m subscipts refer to bilateral port taxes or cuts in import tariffs will shift relaand multilateral versions of e, e, isan index of the tive prices in favor of exportables. These relative nominal rate of exchange between the domestic price shifts should be measured as part of macrcurrency and the currency of trading partner i, oj meso analysis, since they wuil dearly affect the are the trade weights used in computing the various household groups differently. Where multilateral index, and WPP'w are the wholesale domestic price data are good, the effects of such price indies of the trading partner, i (there being policies can be traced directly, with a table simim tding partners). Once calculated for periods lar to Frame 2.2 being computed for the domestic of both destabiization and adjstent, should importables and exportables relative price. It is be cmpared with the diet cakulations of P,/P.* particularly important to attempt direct calcula- Generally, these should given simlar pattems of tions of changes in relative prices within tradables relatve pice change, though the exact movement if quantitative restictions are imposed (and adof theditect pice series may be lagged somewhat justed) on imports, since the indirect methods behind thee-series. using world prices (discussed below) are not 2.49 These estimates of e have been made on generally valid. For such commoudities, world the assumption that the nominal exchange rate ptices are a poor guide to donmstic price levels. (e) is the same for all tradables, so that the index 251 Where price data are poor, the analyst is provides an accurate estimate of changes in the obliged to use world prices and from them to relative price of au tradables. This may not al- compute the domestic price. This procedure is ways be the case. Quite apart from the problem only valid for importables when quantitative reof multiple exchange rates (in which the official strictions are not applied. Assume that s = 0, but valueof ewill vary across sectors), some tradable that k > 0 (see paragraph 42 above), and that the sectors may purchase or ser foreign exchange in g ent reduces k as part of its adjustent the parallel market, where the domestic curency pdre of foreign currency is generally higher. ' packase. The effect on the domestic price of importables can be calculated from, Thus the calculated value of e will prove a nisleadingindicatorofresourceflows. If producers P. e[p,0(1+k)j. of exportables are more liely to sell foreign exchange in the offical maket, if importables pro- With changes in both k and e taking place under ducrs tend to purchase in the paralel market, adjustment, it should be possible for the analyst and if an exchange rate depreiation narrows the to compute a time series for those Inportables 34

43 Labor-nmrkt e ts whose pices have been changd as a reslt. Calculatng he prike d in tjs way nme t 1 2.I: An example of relative-price fle analyst can derive the poicy-induced changes directly. Estimates of AP,, (based on Pm*, and be An ilsatio of the type of pro encmtered In and Ak) can then be used in assessing equation 2. igawpdeeseiesb9obefo lmdes Shilary, computations can be carried out for P.* (1988) _eme of ad4uslmt in Malawi and Tanzausing eaetd ha na For MalawL he had acs to eable data on the 2.52 Although not directly rlevant to theneso- domestic pike of expota but not for importables micro analysis plans set out elsewhere in this Since Malawi did not use demit mport quots or pre volume, the nalyst may wish to establish whethr cont0os he was able to Infer how he wod prim and these relative pricedianges haveindeed induced tariff rate. the domestic pitce of importables. By constrast, Tanzania used both quantitatve restcto and meource transfers. Estimates of changes in out- prica control extensvely, and this procedure could no put (or value added) in the various sectors should be dfsed ty, r Ws nry Corner was provide evidence of whther adjusmnt is suc- able to use sector-spedfic GDP deflator From fese ceeding in shifting resurces out of nontradables set of calculai3 he was able to estimate a prie seris (and possibly previously protected importables), for both exports and importables for the two countriea and into tradables, especially eortbles. More He also computed nontradables pries from vaious direct evidence on labor transfers is discussed in sourcs: the next section. Domestic pres of expot Impt be and nontradbles, i As with price data, the possibilities for AW- N,at Njo analysis of the real wage effects of adjustment are IqWt Ne m- ot im- edcertain to be constained by datavailability. It is E a hudh L p pwtww a" unlily that reliable labor-force uveys will have U173 lod WO ICO 1X been conducted at appropriate ponts for analyt cal purposes (that is, at the start of the periods of destabili adiulsnem and and some time after adjustment polides have taken effect) s Moreover, these Moreover, data will usually refer ondy to ~~~~~~~~~~~~i fomal setor wages, and movements in these are 1981 m s4 liely to be different (quantitatively, and possi bly also qualitatively) from infonral wages Gener y, data on publi-sector wages are sw S :T likely to be more readily available than for other wage senes, and these can provide some clues of Methods of calcduatlo real wage trends overall. Similarly, minimum Malawi: exporta - direct data on domesic export wages canbeused to estimate wage trends change pricesm importable - cim Import pries dmes one plus average tarff rate; nontradabes - simple average of during destabilization and adjutstment. Needless swci fees, entertainment, personal servies, tavel doto say, such wage trends should be expressed in mesic help and low-income housing. real terms, with nominal wages deflated by the Tanzanai agricultural exports - from 1985 Agriculconsumer price index. tur Price Review, Importabls - manufactuing value Ideally, however, measures of wage dif- addeddeiatornontraes-weightedavergeof CDP ferentbls, espeay tween tradables and non- deiators for contuin, wholesale and retail trade tradables, and formal and informal sectors, should and traporl The lesn from thesestatt e be computed, since these are likely to change hdam humtatbeu bato itubmed durng adustent (at leastinthe short nm). Thus, fil price series Note that for Mawawi, the data deary if private-sectr wage series are available, the show that nontadables beame markedly chea reaanlyst should compute changes in the public tlive to importables, mandy as a result of expenditure sector wage in relation to private-sector (nmrket) swtching poides. In Tanzana, the evidence indicates wages throughout the period under study. The ltatpricemovm tsencdagedresouettnlsfersout expectation is that during destabilization, the of exportables into impta and nontaddbs public-sector wage will have risen relatively, and. 35

44 fallen during adjustmnt. Siniflar patterns ight market to obtin their credit requkemets. Sinhe be expected of the rural-urban wage differential. nest charges are fixed (and generally low) in An analysis such as this can reveal how relative the organized marke such borrowers face inwage trds have been changed as a result of crased interestchargontheirnewdebt. Interadjustment. Themaindatal mltatlonsarecertain est rates In the kerb market, which are flexdble, to be in the area of the informal-sector (or mar- will theefore rise as the credit contraction in the ket-detemined) wage. It is aey unlky that foma arket pushes numre borrowers into the such infonration will be available over time to kerb market. Thus the effect of monetary conpermit meaningf macro-meso analysis traction is to rstrict th avaiabty of credit in 2.56 Even without informal-sector wage data, the orpnized maket, and to increase interest officdal wage trends can be compared with aver- rates in the kerb marke age levels of living for the popultion in general. 259 In some program4, the fixed-interest re- An example of this is found in Collier (1988). gome of the formal market (refred to as a "re- Using national accounts estimates and the CPI, psion" of the money market in the literature) he computed real private consumption per cap- isdismantled, so that Interest charges are allowed ita for Malawi and Tanzania. At the same time, to settle at their market-clearing values. Thus, in he esimated a real-wage series for the fomal addition to any decrease in credit availability in sector from official sources, again deflating using the formal market, bonowers may face increased the CPI. By expessing the real wage relative to interest charges in the market as well. average real consumption, he generated a time 2.60 Once again, the extent to which these series which showed for both countries that the changes can be convincingly analyzed depends real-wage/real-consumption index fell from 100 critically on data availability. Information on in 1973 to around 65/70 in Thus he was cedit and its tem si genery available from able to show (albeit through an Imperfect indica- the official banking system, but this can prove tor) that formal-sector wages had fallen in rela- totally misleadig as a preparation for the nesotion to peasant living standards quite dramati- micro analyses of most households obtaining their caly. credit from the kerb market. It may be, for ex- 257 In the absence of wage data, there is some ample, that interest rates are not increased sigsopde for using the resiuts of the SDA Integrated nificantly among the larger banks, but because of Survey to assess the direcdon of likely wage credit raint, kemarket rates increase. Uschanges, at least in the short run. This derives ing offkcial rates would be inappropriate in anafrom the fact that the real wage will rise (fall) as a lyg houseods whichborrow through the kerb result of an expenditure switching policy if market. Data on the kerb market therefore are workers consume mainly nontradables essential. If possible,therefore, nominal and real (tradables). An examination of the expenditure rates of interest applying ir the ornized bankpatterns of wage-ewadng househods should en- ing system and in kerb markets should be comable the analyst to infer something about the di- puted for the period of destabilization and adrection of the likely short run real-wage change. jsent. In this way, some historical estimate of the effect of adjustment on the credit market can Changes in cedit market beobtained In most African countries fiscal contrac- Expenditure-reducing effects tion is closely associated with monetary contracbtion, so that adjustment is usually associated with 2.61 The first stage in the analysis of the effects significant changes in the credit market Credit of expenditure-reducing policies is simply to markets invariably consist of a formal market, Identify which of the broad categories of expenwhich is dominated by the organized, moden diture were subject to cuts (and also, by te same banlg system, and an informal or "kerb' mar- token, to identify those items which expanded ket. The former is directly subject to the restraints during the deabizing phase). Expenditures that are imposed under nwnetay contraction, should be dted initially into the four while the latter is affected ondy indirectly. Typl- catewries of equation 9: private consumption and cally, a credit squeeze will reduce the supply of investnent, and public consunption and investcredit in the organized banldng system, so that ment. 5 By dividing thirough equation 9 by Y many borrowers have to shift to the infmal (GDP), we can decompose hanges in te trade 36

45 deffcit/gdp ratio as fouows Frame 2.3: Decompositlon of changes In the traw-deficitgd? ratio by expenditure item AIXO-M)/GDPI = 1- A(C/GDP) - AI/GDP) (pom4) -A(G,/GDP)- (G,/GDP)(11) Again, a table such as Frame 23 should r, &-MJGDp A(=,GD1),%AGDP) A(CGDP) A,G/cDp) prove extremely useful in any prelhminary assessment of the implications of destabilization io and expenditure reduction. It may be, for ex ample, that the expansion of the trade deficdt/ 1981 lod GDP ratio during the period of destabilization jgsg l00 (assumed to be in our hypothetical case), was associated mainly with an increase in government investent, whilst the contraction of the Mtoedumt. deficit during adjusent was due to cuts in pri vate consumption andrinvestment. Such asym mefty in expenditure changes over the cyde of destabilization and adjustment are important to identify, since they have a direct bearing on the social dimensions of adjustment. If the macro The whole table might be prepared for recrent imbalane was caused by an over-expansionary and developnment (or capital) expenditures, and fiscal policy, whereas its correction was borne comparisons made. mainly by private consumption, the social costs 2.65 The analyst should be careful to define an of the adjustment are liklly to be significant. appropriate price index in generating a time so-.63 Given the central role of the government ries of expenditures in real terms In most cases, in the provision of infrastructural services the consumer price index (CPI) will be all that is (through both investment and re-current expen- readily available, but in others, alternatives may ditures), cuts in GC and Gi will be of particular be tried (at least for the sake of comparison). interest for macro-meso analysis However, for These would include a wage index (on the asour purposes, these expenditure items are sim- sumption that labor is the most important cost of ply not sufficiently disaggregated, since the meso- many public services), the GDP deflator, or that micro analysis plans require somewhat more re- part of the deflator which refers to the service/ fined estimates of changes in infrastmctural public sector. The proportion in total real govservices. Chapter Five on health, for example, ernment expenditure of each of the items deexplains health status of individual household picted in Frame 2.4 should also be computed, members in terms of changes in health services and dhanges over ime observed. available to the household. It is dear that SDA 2.66 The data in this table merely show what macro-meso analysis must disaggregate further has happened in real terms to expenditures on the changes in govenmment expenditures during each of the items. Again we have the knotty destabilization and adjustment, and attempt to research task of quantifying the change in infrasestimate how much the various items of expendi- tructural spending imposed as a result of structure have changed in real terms over the period, tural adjustment, since it is on this basis that the and especially during the adjustment period. first difference equation (2) is to be computed. In 2.64 At the very least, therefore, a time series makingjudgements on this, the analyst must bear should be computed for each country, showing in mind the following. the changes in public spending by major cate- * Takingdeviationsofexpendituresfromatime gory. For this, a table along the lines of Frame 2A trend is an inappropriate method of assessing the (overleaf) should be prepared. The expenditure impact of adjustment on that expenditure item. categories depicted in the table ought to be con- As we have been at pains to show, the period sidered as a mindmum disaggregation. It would prior to the adjustment is unlikely to be typical of be helpful, for example, to distinguish between any long term trend, being one in which expendithe three levels of education (pnmary, secondary tures are ;lkely to have been unusually (and unand tertiary), and different categories of health sustainably) high. expenditures(prlmaryhealthcare,hospitasetc.). * Changes in the real expenditures on each 37

46 Frame 2A: Real govenment expendit by stow Soxcial infustnwf we Ewnmk infesbutar Sowd H cjmmi- &wmk Dow EdaWim lh sawh w4hi'e Roads Ueil Was ation sevc itan Tota lg Memorandum items Pentrag chane l a. hdudinglsemaswhasptlion mndeisloese vin ese eaybe lised septyf b. Indunggrlmit dseter wsas. item may not nmssrly reflect changes in the Economy-wide modeling techniques avail~bility of services to households. This is because cuts in expenditures may be associated 2.68 The analysis suggested above relies on wvih more efficient use of resoumres, whilst the an examination of the historical record. Tbe key opposite nay be true of periods of expmnsion. ies in identing periods of destabilization and The effects of xpenditure cuts wll vay by adjust Thisthenpermitstheanalysttodraw relon dependin on how the cuts are applied conrlusions about how adjustment has affcted within each sector. For example, expenditure markets and infrastrctpre. But this type of cutsin higheeducationareunlikelyto haveany analyslsissbject toanumberof seriouspitfalls. real effects on education services in rural areas. First, the historical record does not Imply aus- But if the cuts were applied 'mainly to primary tion, since what is observed happeniig over time schools, rpral areas nmay suffer disproportion- is influenoed by a wide range of factors, and not ately. Smilarly, cuts in teacher recruitment will simply the macro-iralaces and the conrecve inevitably harmn rural areas more than urban, adjusmt policies. Movements in laive prices, given the prfence teachers attach to working for example, might reflect changes in supply in urban areas. The quantity adjustmt ofsucha eonditions that are entirely unrelated to adiustpolicy would be borne mainly by rural areas. ment-including the favorabl effects of techni- * Some changes in government expenditures cal progress and the adverse effects of drought. may have taken place quite independently of The mentodology suggested above canwt enadjtmnnt, but are merely associated with the tizly separate out these factors.' adjustment program chronologically. Obviously 2.69 The seonid main problem with the sugthereis little that canbesaid by way ofagenerali- gested analysis plan Is that it takes a frst order zation here- it is simply a matter for the analyst appwximation of how policy interventions affect to make a judgement in his or her country. markets and infrastructue To gain an under- 2.67, However, the memorandum items in standing of this, consider again the digamnatic Frame 2.4 are designed to identify the key chro- presentation of the plan in Figure 2.1. The arnology of the expenditure chanes, again divid- rows show how the rationale of the analysis pring the period into naning sub-periods which ceeds, begning with the effects of policy interreflect destiization and adjustment., ventions on markets and infrastructure,tadthen and are ondy iustrative exaiips. These examnipg the effects of the nmso variables on would need to be spedfied in each case. households. This approach ignores (or assumes 38 : ~~~~~~~/

47 as insigificant the resons of households to the SDA analytical program. This is no+ meant, market and infrastructural changes - so called however, either as a review of such models, or as igher-order effects. These are shown in the guidelines for the construction and use of moddiagmm as the dotted line, linking household els, since there are obvious space cdnstraints in a responses back to markets. Thus, for example, a chapter such as this. Our purpose here is to policy which raises the market price of a comn- highlight the strengths and weaknesses of the modity will lead some households to purchase approach, and to identify the main features an less of it, thereby reducing demand and causing economy-wide model must possess to be of use the price to fall back somewhat This might be in the SDA analysis program. related to the time period that is allowed in trac At the outset, it is important to emphasise ing the effects. The first order effects are impact that a model of the economy is simply a piece of effects, with the full effects taking time to work capital equipment. And as such, it is intended to their way through the economic ystem. The first aid our thinking - not to substitute for it. Econorder impact on prices may well be greater than omy-wide niodels are to be used in a process the full effect, since the initial change in prices is which is human-capital iniensive. They are to be moderated by demand and supply responses. combined with the application of human judge- Notice, the feedback loop is drawn only to mar- ment, and not to replace it. In this way, the lser kets, since this is where higher order effects are to will not be tempted to react in an extreme way to be expected. There may be ways in which house- the model - either to dismiss all its predictions hold responses may cause further changes to in- as too mechanical and useless, or to accept withfrastructural services, but these are likely to take out reservation anything that is produced by the some tme to have some effect (as with, for ex- model (simply because it appears to be 'scienample, responses through the political process). tiflc'). The true response to and use of the econ The third weakness of the analysis sug- omy-wide model Involves an application of the gested thus far is that it ignores the interactions analysts own judgement and understanding of between markets and infrastructure. Yet we know the country to which the model is applied, and to that the economy functions as a system, with interpret the predictions of the mnodel accordchanges in any one market influencing what ingly. happens in other Thus, for example, an in The main strength of the model lies in its crease in the price of food will increase produc- formalization of the thinking process- it makes tion of food, which in turn will raise the demand explicit what is often assumed implicitly in less for farm labor. This may raise rural wages. These formalised logic. It also serves to simplify the effects need to be traced in a systematic manner. real world, by ignoring minor detail, and focus Fially, the above methodology does not ing on the more important interactions. In this produce any robust answers to the question, What sense, a model is a small, inferior version of the might have happened had a different set of ad- real thing - in this case the economy. Many justmentpoliciesbeen applied? The counterfactual models are important bridges between economic case is not readily derived from the historical theory and the practice of economic policy (Derapproach that we have suggested. Yet, in most vis, et al., 1982). Whilst theo l models are an cases, goverments would like to have some idea essential starting point in considering a real-world of how alternative policies might have worked problem, they cannot take us vtjy far. This is (or not, as the case may be), and such impres- because at best they can only provide qualitative sions ae difficult to form without more struc- predictions, whilst the pa'icy practitioners must tured and rigorous methods. make quantitative dedsions Theory may suggest that a devaluation policy will work under The strengths and limitations of modelling certain circumstances, but the practitioner must know by how much to devalue. But theory isnot 2.72 It follows, therefore, that a more system- always clear about its qualitative conclusions. atic approach to macro-meso analysis might be This occurs especially in cases where there are needed. The use of economy-wide models has two opposing effects of a change, the net result already a tradition in development economics depending on which of the effects is the stronger. and policy design. In the remainder of this sec- For this theory alone cannot provide an unambition, we shall review the potential of economy- guous answer. In such ircumstances, the model wide modelling for nacro-neso analysis in the comes into its own, since it can simulate the two 39

48 opposing effct, and reveal which is the stronger. such as monetary and interest rate policy. Te Many policy issues are of this type, and require macro-meso model must, to some extent, be able both quantification and simulation before an an- to track changes in both real and monetay polcy swer can be obtained. instruments at the aronic level. This 2.75 The main weakness of modelling les in its issue is discussed further below. highly structured approach. First, it assumes * Since many adjustment policy instruments that the economiy can be described through a set have ther prdncipal effects through changing the of relationships which remain stable over time. incentive structues of product and factor marnsohfar as these might change, the predictions of kets, the model should treat product and factor the model wil prove unreliable. Second, econ- prices as endogenous. This means that product omy-wide models are generally not designed to and factor supplies and demands must be modtake into account qualitative changes in the econ- elled, and price changes related to excess deomy, which by their nature are not easily quanti- mands. Price- and wage-endognity, however, fied. Finally, the degree of disaggregation that is may be constrained by rigidities which prevent mearnngfully feasible in an economy-wide model m etsfo clmaringinstantaneously, and which remains limited - not by the capacity of the result in ratioing i the marwkets and persistent computer, but the understanding of the user. If excess demands. This would obviously depend the model is too detailed, it quickly becomes a on the characteristics of the markets in the counblack box to the user, resembling more the vast try conced. These constaints can play a pivcomplexity of the real world than a simplified otal role in explaining how adjustment affects version of it to aid our thinking. maroets and thereby household incomes. Wage rigidity in particmbar, can determine how the la- Required features of maro-meso modeds bor market transnits the effects of adjustment policy to households These considerations apply to models in * Finally, the macro-meso model should ingeera but what are the specific requirements of dude the household sector, and trace how houseour F sent concern, the macro-meso analysis in hold incomes and expenditures are determined. the I *A analytical program? The main features We have already stated that the level of houseof ait economy-wide model to trace the meso hold disaggregation that is feasible in an econeffects of adjustment policy may be summarised omy-wide model may be too limited for many as follows: policy issues. And although the main purpose of * The model is to be used for policy analysis. the model is to trace meso-economic effects (at As such, it should be designed to aid our under- least in the context of the SDA analysis scheme), standing of the way the economic systen oper- the household sector remains an important inates, rather than forecasting the future. It must gredient in the model. This is for two principal therefore treat policy instruments and targets reasons. Frst, the model can provide useful inexplicitly. It should also be able to trace the sights into househowd effects, even if this is not causal mechanisms at work, linking the instru- altogether sufficient for policy makers. Second, ments to the targets. 7 Among the tagets should without the household sector, the model would be incduded the fiscal and extenal deficits. be unae to tbace the full nnmo-economic effects. * It must involve a minimum degree of disag- Figure 21 showed how househd responses play gregation, at the very least distinguishing be- an important part in determining what happens tween the main producing sectors of the econ- in markets as a result of policy interventions. omy: exportables, importables, protected im- Therefore, the detrnation of household inportables and nontradakles. A macro-model is comes and expenditures shoud be considered as inappropriate, since it is incapable of analyzing an essential element of the macro-meso model. the critical strucnual shifts brought about by adjustint policies; Do AGE modes meet the bill? * On the other hand, our ideal model must be capableof representingmaro-policyinstrumens 2.77 Applied gral equilibrium (AGE) mod- Whilst many instrumcnts are 'real' economic els are frequently used in addressing policy isvariabes, such as government demands for con- sues in an economy-wide fmnework. 8 The main sumer and investment goods, and many price features of AGE models is the gal equilibpolides, other policy instruments are mwnety, rium nature of the model. Producing units are 40

49 represented as purchasing factors from house- general rule, the choice of numeraire in such holds, and therby generating value added. In models is entirely arbitraiy, since the model is return for factor services, households receive in- not intended to explain the absolute pice leve. come payments, which are allocated to savings This means that such AGE models are incapable and consumption. The sum of these production of tracing the effects of macroecononic policy, and consumption decisions yields supplies and especially when such policy intervention is 4emands in the various product and factor mar- through the operation of monetary variables kets These are reconciled though relative price 2.1 AGEmodels,however,canbe modified t adjustments, which yield zero excess demands in incorporate a macroeconomic superstructure. the markets unless some rigidity is assumed to This can meaningfully interact with the multiexist. Descibed in this way, the AGE model sectorgeneral equilibrium model, and enable the seeks to trace the operation of the real economy. user to conduct macroeconomic erents in It yields predictions of what is likely to happen to an AGE setting. Bourguignon et al (1989) have the structure of relative product and factor prices made an important recent contribution in this if the assumptions of the model, or if policy in- area. Their refinement of the AGE approach instruments such as trade taxes, change. As such, it volves the addition of a macroeconomic model to is an empirical application of orthodox real-eon- the general equilibrimn framework, in which the omy theory, and as such, t describes a barter macroeconomdc sub-model yields the level of economy. aggregate demand in the economy (through a 2.78 It should be clear from the above that AGE variant of an IS/LM system), whilst the general models show considerable promise for macro- equilibrium sub-model computes the real-econmeso modelling. They explain what detenrines omy side, yielding aggregate supply. The price rlative product-and factor-pice movements, and level is given by the relation between aggregate showhowresourceaflocationrulesoperate. Most supply and demand. Their macro-model con- AGE nodels utilize base-year Social Accounting tainsgenralasset-marketequiibumconditions, Matrices (SAMs), which represent how inconws whichdeterminea vectorof ratesof return. These generated by production activities are mapped in turn influence the level of aggregate demand. across to households, and how expenditure deci- In a simpler version of the same, Demery and sions by households are allocated to sectors. As Demery (1990) take two financial assets (money such, AGE models have a built-in focus on in- and bonds), and trace the link between the monecomes and their distribution. Finally, AGE mod- tary and real economies through the interest rate els usually explicitly model the external sector, in textbook fashion. and can be used meaningfully to analyze the 2.82 In other words, there are now AGE modeffects of trade policy. Such qualifications make els in use which do incorporate the macro-econthem eminently suitable for SDA analysis. omy into a multi-sector general equilibrium 279 AGE models, however, have two princi- famework. It Is perhaps too early to judge pal weaknesses from the macro-meso analysis whether these models provide an empirically pepective. First, they do not usually embrace convincing description of how macroeconomic the macro-economy, especially its monetary changes influence the real ewnomies of developcomponets. Second, they only trace the effects ing countries of policy interventions on markets, and do not treat nfasuctural effects We shaldisss each 2.83 AGE moomss AND NaA5cIL The seclimitation in turn ond major limitation of AGE models when applied to the analysis of the effects of adjustment 2.80 AGE MODES AND he MACEoNoMy. AGE concems their inability to capture the full effects models are principally concerned with relative of infructural changes This applies as much prices, resource allocation and trade. They are to economic as to sodal inftastructure. If adjustnot usually designed to analyze macroeconomic ment entafls cuts in govement expenditure on changs. Most AGE models follow faily simple education, health and physical infrastructure, price nonnalization procedures, whereby the there are two broads effects. First, there will be aggrepgate price level is taken as exogenous, and income losses for those employed in the sectors all absolute prices are simply ressed in term which previously benefited from these expendiof some numeraie Relative prices, of course, tures (ie. teachers, doctors, road construction remain unaffected by such nornalizatiom As a firms, and so on). These direct primry-income 41

50 effects of expenditure cuts can readily be mews umst nenuity (if not imagination) to push the ured by AGE models, simply by tracing the full available data as far as they can convinrdngly go. genal equilibfium Implications of the resultig 2.87 In cases where economy-wide modelling contraction in sectoral demands. 9 To do this, the is not feasible, the approach wecom nded here cuts In govenment expenditure Items need to be concentrates on the historical record. The analytranslated Into sectoral demands, so that they can sisprogram that we have recommende depends bereadable input into the AGE model (or a SAM). on the identification of periods of destabilizaton 2.84 The expenditure cuts, however, will have and adjustment, so that changes in the mewso indirect effects on the incomes of households who economy between these periods can be attributed previously benefited from the infrastructural (for the most part) to the adjustmt prgrm. seris that were produced. Cuts in expendi- This is an imperfect solution, since we know that tureoneducationwillnotonlyreducethepnmazy there may well be other factors which have a incomes of those engaged in the education sector, bearing on the historical record. The analyst has they will also adversely affect the education of to use hisown judgement in maklngadjustments those who previously benefied from the serv- for these extraneous influences. ices, reducing current levels of household welfre, and future levels of income. The same can Notes be said of cuts in health expenditure. A reduction in otlw infrastructural services, such as road 1. Using IMF minolog. consruction, will have similar effects. House- 2. This abo assumes that the internaional terms of trade holds who cannot gain access to markets and (PIP.* remain unchanged. Inputs will exrience declines in output and in- 3. Weignore here theuseof money wageindices as deflacome. These indirect effects on incomes and tas,thoughsehavebeenusdeffeivelyespecialyen welfare cannot readily be tced in an AGE envi- indices of inteanal competitiveness are required. onmient. 4. Edwars989:105af)cnputesrealeffectiveechage rates using both official nmnal exchge rate and rates Conduding observations which apply In parallel foeign exchange markets. He found that tey bore very little rdation to each other. Indeed, in Macro-meso analysis in the SDA research out of the 28 cases considered, the conrelati coeffident scheme is certain to be the most tasking of all beween the two hidic was neptve. Tis suggests that those outlined in this Volume. Yet this work is an where parallel markets are important,. the use of the offidal essential prerequisite for what folbws Without rate can be serioudsy mwsleadng an understanding of how adjustment policies in- 5. Tis leveld of tsagpgtion Is chsen beue Is t confluence the main explanatory vanables consid- dsset with what is available om the nationalacunts. As ered by the meso-micro plans as important, it we shall diswe, a much more rellned classification of exwill not be possible to link the results of the SDA pendiures will be requir in assessing the inrastructural data gathering initiative with policy design. implications of adjustment What aan be done in analyzirg the meso 6. Though It does attempt toseparateoute effects of the effects of adjustment policy depends on the 'recession' fum of the adjustment by cely subavailable data in the country. We have tried to dividgthe histoic record. keep the data requirements to a minimum in 7. FotIng mod usally rely hewanlagged enwhat we have recommended here. But in the dogenow vaiabe ad rnduced fom s ons. Allast analysis, the data wnll constrain what can be thugh they yield better ft, such models geerauy mask the causgal medaism nvolved. 8. For an excllent summary of the ma features of such model, see Dervis et al (1982:Cb. 5). 9. By the same token, a SAM an perform sinmla ments. horbede et al (199W) ae the pdmary-income dis- tribution effects of alteative fwal poiis in Indesia us- inga fix-prtce SAM mulplier model achieved. At one extreme, we have presented some fairly basic analyses requiring only national accounts data. On the other, there is the application of AGE models, with their potentially voracious appetite for numbers. It is probable that most countries will lie sonewhere in between these cases. In all cases, as some of our illustrations have shown, analysts wil need to use the 42

51 The poverty profle Intoducton and objectves viaton is a necessary condition for sustainable long-term growth, for wiout it, humn capital 3.1 This Chapter is contcr with poverty, accuuion is threatened. Thid, it Is a matter and how it is Influ nced by struc tu adstm of some uen for most poliy makes Identipolicies It sets out a plan of analysis which is fying the poor andaesing how they fate in the intended to guide retels and policy nmkers light of policy reforms can be crical In some in utlizig houseold data in developing coun- extreme cases, adverse poverty efft can cause triesforthepurposeofodrucdtipovertypro- politica instabiity, and it is vital for govenfles It is directed particulrly to the analysis of ments to be able to make some ex ante assessthe SDA Integated Survey. Te methods pro- ment of poteni pverty problem arising from posed are rlatively dmple, and ost of the re- their poliy Interventionsn sults ate designed tobe presented in tabular form 3.03 The porty proflle hould also be given so that they can be understood readily by deci- chronobgia priority in the SDA aralysis prosion maker. As a profile, the analysis of this gra, swnce Its rests are certai to be essnial Chapter does not preted to prove causation be. inputs into other areas of analysis This MCapter, tween adjstment and povey, but to indicate therfore, is best consdered as an ubl' unsom inferences about t Hlikely effects. Neither der which the other Analysis Plans ate plaed. does the profie enter Into anaysis of the causes The poverty prfile shoud be the first of the of povrty. Again, it may provide some insights analysis componets to be undertakn in each i ths (as, forexa; through the muli-van- conmtry participatiin the SD)A pogram. ate anls of Incom/ependitue), but ths is 3.04 The main objectives of the povety profi's not ft main purpose of the Chapter. The analy- awe to: ds presented here is meat to be a first cut at * identify which are the major povety groups Identifying which groups are poor, and in mak- in the onmtry conceaed; ing sonw me _ of how thy might be af- * assess how these groups have ben affiected fected by ptrucuadjdustment. by structural adjustment proan; and 3.02 Oher Chapters address different aspects * drawsomepolicyinferenoes,aonmingboth of the social dimeso of adutmet, but the the design of adjstment polces and the aevi povet profile must be given priority for a nur- tlon of poverty. ber of reasons hst, the goal of poverty reduc- 3D5 The first objective, though Itself a major tli is paramount It is one of the most finia- empic undertalkig is needed as a bads for mental objectves of devpnent, and one whih cgen the sewed and thidd,more furndafminitselfdemandsattentlon SewcTnd,povertyalle- tal objectves (from the SDA pmpective). The 43

52 difflculties of establshing emplcaly a causal order to proceed witthis, two kyquesyions link betwen omlc poicy change on must be resolved. First, what mmure ofwfare Is the one hand, and the niodnomic outcomes to be adopted as a basis for poverty anarlyss? at the household level on the othe, have aleady Second, how hould poverty in the community been discussed In the companion volume, World be measured? 1he second issue is dealt with in Bank (1990), and they are addressed in Chapter the next section. The first has been disussed to Two. Care mst be taken in the analysis of pov- some et in World Bank(1990),although some erty and adjustmt to avoid simply traing pov- amplification would be in order here. Idedly, we erty historically over the adjustment period, and should use some mease of indifidual welfare, but assuming some causa link. In order to ensue tis is not always feasible given data cnstints. that the analysis goes some way in establishing The data base addiressed by these plans is the the causal mechanism between maco-interven- SDA Integrated Survey (S). IS data do not permit tions and microutcomes, a key anytical de- Individual-level analysis of expenditur& Howvice is introduced - the meso economy. By ever, this does not prevent the analyst from deundersnding fist the link between macro pol- veloping data constucts which rflect individual icy interventions and meofcnomic outcomes well-being for example through expressing (in terms of markets and soial and economic householdlevel variables on a per capita bais. infrastucture), and secondly the effects of meso- 3.8 The main problem faced in analyzing ineconomic change on the household groups, SDA dividual (or household) welfare is that utility is analy8es will enihance understanin the proc- neverobserved directly. Whatisrequiredissome esses thrugh which households have been af- counterpart td welfare which can readily be obfected by policy interventions, which in tun will served. A strong theorec case can be made for icrease confidence in ascrbing causation in the using expenditure as our proxy for welfare. 2 Asrelationships observed. The poverty profie as- sumethatindividualsconsumetwobroadcasses sumesthatsomeunderstdnghasalreadybeen of commndities - ihose tfhat they puhase in gained about the key macro-meso effects of product arlkets, and those that they produce strctural adjustment, as explained in Chapter themselves. For such individuals, the problm is Two. In particular, it assumes that the effects of to: adjustment policies on relative prices, economic structure, growth and sodal Infastructure are Maximize UM, xi) xeaswnablyq weo undersxx ds Subject to ptx Sp 2 (q6-x2) + A-C (1) 3.06 The next section isdevoted to adiscusion of measures of wefare which are to be used in where x, represent vectors of market-purchased thepovertyprofilse Paragraphs3.19through3.33 goods ( = 1) and the cnsumption of home prothenreviewthe nof povertyand dscribe duced-goods (i = 2), pi are the its m u e Paragraphs 3.34 thrugh 3.77 prices, q6 is a domestic output vector, A other address thecentralconcernof thepovertyprofile (exogenou) income and C total production in- - an exmination of how adjustment podices put costs The values of q and C are preeeteraffect poverty. The thodlogy outlined here mined through the household's production (netemphases the first order effects of adjstment rvenue-maxmin decisions.' polcy, whreby adsmnt affects mwkets and 3.09 The dual epression of (1) represts the nfrastructure and these affect poor bousehlds. hidlvidual'sobjectiveas minimzlngexperditure Higher order effects are conidered ondy briefly. subject to some utility constraint: Basic needs dimensions of poverty are disussed i paragraphs 3.78 through 3.85, whlst para- Minimz px, x pa graphs 3.86 through 3.91 trace some of poverty's dynamic aspect. Someconcludingobsrtions Subjecto U (x,x 2 ) U (2) are made in paragaphs 3.92 and where U is some minimum acoeptable level of Measuring welfare utility. Solving (2) gives the expendihurefsction: 3.07 Webegitwithourfirstobjective,whichis ErE[(p 1 p 1 ),Up (3) the idendfication of the poor in the countiy conced, and teir madn economdc activitie In Expenditure is a monotonicaily increas func- 44

53 tidn of U, so that if ail individuals ihae the same hold survey ltself. preferences (or utfity funion), and if tey face 3.11 SImilarly, if compardsoks of expenditure the same prices (both explkit market prices and are made over me, account has to be taken of Implidt own-prducto prices), the raning of changes tn the oveal price level. Idealy, reexpenditures will be the same as the raning of gloa Consum r Price Indices (CPUs) suld be utilities. This then isa sound theoretical sifica- used to account for changes in price over time, tion for the use of expenditure as a proxy for but these are rarey avaiable. In practice, ondy a utility in poverty analysis. It is based on the nadonal CPI will be available. The use of thds assumption that individuals are utlity maxinmz- index is valid so long as there are no significant es, and that the main arguomts of their welfare variations in the rate of inflation among the r- function are the conunodities they consume. glons of the country Theuseof expenditureasourhinicatorof 3.12 Within this frantework, a number of inweare also assumes that individuals face the come- and expenditur-based indicators present same set of prices. In the real world, however, thenmeves for consideration. These will each they may face different prices - of both pur- need tobe computed from the raw data collected chased and own-produced commodities. In such in the survey, since they are not obtained from cases, it is not valid to use nominal expenditure the respondent in a directly useable fom. The as a welfareindicator, simply because some vari- following is a short-list of possible measures of ation in expenditures will be due to price differ- wlfar ences. Tlus, if we are comparing expenditures * total household income across regions, care must be taken to account for * total household expenditure rgional price variations in computing an expen- * per capita household income diture variable to reflect relative welfare levels. * per capita household expenditure To do this, a reference region needs to be chosen, * tota household income per adult equivalent and cost of living coprison made with other * total household expenditure per adult regions. A suitable index for deflating regional equivalent expenditures into the prces of the reference re- * per capita food consunmtion gion is the Paasche cost of living index. The * proportion of household budget spent onfood Paasche pice index (I)for rgion I is given by, 3.13 These measmes do not always produce the same results, and so it is important for the i ( q >p~) / ta4) ) analyst to be sure that the indicator selected is P -. qapu pu, supa appropriatefor the purpose in hand.' The debate over whether welfare is better measured by inwhere p, and p, are te pies of conmodity i (i come or expenditure (discussed in Grootat =1,...n) in the reference (0) and non-reference (1) 1983) in part derives from the empirical prbregons respectively, and q6 are the expenditures lems that are encountered in nasuring income on coodity i in region 1. The index uses the thtough household surv. In general, housnon-rerrience region expendites as weights t hold surveys give expenditures In excess of incompute the regional price index. Altematively, come for many household groups, which raises (4) may be written, questions about the reliability of the income estin n nates. Our sugestion, thefore, is to adopt an I / =IV; ( / Ie exp,,) ure-bas wela measure -rotal Ex- " ffi penditures' should be the sum of all monetary where V. is the value of u emft in vgion I expenditures miade by the houseihold, c wmp on commodity i. Usng this index to deflate ex- ffon of own-producdion, and the Imputed value pendtu across regis will yield a LaspIr of sevices derived from the owrship of conreal expenditure index, in which expenditures in sumer durables and housing. Expenditure data eah region are valued at the prices of the refer- must be corected for regiona price variatons, ence region. These computations wm require a usingpaaseregionalpriceindicesasdiscussed. reliable data set on pices across regions in the 3.14 Our main inerest lies in individual welcountry. Price data may be derived from the fars But since the ISisdesignedtocolectexpecoununity urvey or from other sources, and ditures at the household level only, some adjustexpenditures (used as quantity weights in the ment is required to the household-evel indric- Paasche index) can be derived fom the house- tors to reflect more accurately the well-ing of 45

54 the individuals within it In order to account for Age of dkl A" qubiw sok differenres in househld siae, total household expenditure whould at least be measured on a per 7, 12ym03 capita basis. Altematively (ad preferably), addt equivac scal hould be used to take into ac- 218yeas 1.0 count differenes in the age and gender structe of households, though this will inevitably make Even if analysts do not have the opporunity to the analysis more complex (see Deaton and compute their own country-specific weights In Muellbauer, 1980: Ch. 8). the pov profile, they nuy wish to use weights 315 The reason why adult equivalence scales such as thse, and compare the results with those are preferred arises from the fact that children using simple per capita expenditure. are generally considered tobe less demanding, in 3.17 These mon-meriec measures of welfwe the sense that an additionwal child requires fewer shoud be supplemented by other indicators, as additional expenditures to maintain household we shall see below. Our rationale for using an welfare than would an additonal adult. Simply d e welfare ndic atorassum ed that dividing total household expenditure by house- only commodities entened the welfre function. hold size (the per capita measr) will give a There is nothing in principle against exending misleading impression of welfaredffrne For this to include otherelements in the welfare hcexample, compare two households with the same tion, incduding various basic needs that are not level of total expenditure and of the same size, obtained through market purchases (such as only with one comprising all adults and the other health and education). The main problem arises consisting of an equal number of adults and chi- in selecting shadow prices at whih to value tese drer. Clearly, the household with the children seies. Our prefernce, therefore, is to treat wilt enjoy a higher level of welfare, simply be. health and education separately (see paragraphs cause its child nmmbers do not require the con through 3.85 and also Chapters Five and Six, sumption lev of adults. respectively) Dividing total household expenditure by 3.18 By the same token, some nay argue that lousehold size implicitly ascdbes a weight of the welfare function of the poor does not simply unity to all household members, irrespective of consist of the comnwdities they consume, but of their age. Applying an adult eqivalent scale more fundamental quality-of-fe determnants. means that some nmes will be assigned a For example, for many poor households, welfare weight between zero and unity, depending on may depend simply on whether there is suffitheir age. Thus, a child between the age of thir- dent food available, or whether ildren are sufteen and seventeen might be given a weight of ficiently nourished. A case can be made for eon- 0.5, implying that his or her consumption re- sidering other indicators of poverty which may quireenbt are approximately a half that of an reflect this -such as food consumption or measadult Sinilarly, children under seven years of ures of nutrition. There are other arguments for age might be assigned a weiht of 0.2 on the using food expenditure only as the welfae indiadult equivalence scale. Thus, instead of divid- cator. Such expenditures is tess suseptible to ing total expenditure simply by the number in economies of scale (thus not reqirng e use of the household, eah household member is as adult equivaence), and they are easier to meassigned a weight depending on age, and the sum ure accurately, sinc respondents tend to recall of the weights is used to divide total expenditure. more clearly frequently purcased Itens. More- The procedure to construct adult equivalence over, the use of food expenditures does not rescales is beyond the scope of this chapter. Inter- quire any conmlex imputation of expnditures. ested readers are directed to Deaton and Muell- Analysts may wish to adopt a food expenditure bauer(1986) for furtherdetails. Theissuethatthe neasr of welfare in addition to total expendianalyst must face here is whether to use weights ture, and make compariso of the results.6 In any derived from extemal sources or to attempt their event, these considerations are taken up In other estimation using IS data. In his analysis of wel- chapters of this Volume - especiay Chapter fareincotedw'voire, Glewwe (1988) uses weights Seven on food security and nutrition. We shall, which we derived from Sri Lank and Indorei. therefoe, concentrate our attention on total-ex- These are given as: penditure based welfae measures. 46

55 The nature and measument of poverty 322 Calculating an absolute poverty line is usualy fraught with both conceptual and en If it Is agreed that the indicator of house- piricambiudties. What is the minimum caloric hold wele is to be per capita household expn- intake and leve of food consumption that is rediture in most cases, this leaves the seconld key quired to avoid maklutrition? What is the conequestion that will need to be addressed: how to spondig income required to purchase this food? determnw the extent of povery within the con- What non-food commodities should be considmundty? As poverty and its alleviation is at the ered as essental for the commodity basket which center stage of the SDA endeavour, data analysis defines absolute povty? These calculations, must adopt measures of poverty which wil be which wil of necessity utilize the IS findings helpful in guiding policy makers and in faciitat- themselves, will inevitably involve the personal ing links between poverty and the main struc- udgenent of the analyst, and a measur of arbitural chaacteristics of the enony, which are triness. If an absolute poverty line is to be themselves likely tobe influenced by adjustment calcuated and used, ft is important that the asprogms In any measure of poverty, two broad sumptions on which It was calculated (the basket issues present themselves first, the identification of goods on which it is based and the vector of of the poerty lin; and secondly, the choice of a prices which is used to translate this real bundle single index to measure poverty. into income terms) are made explicit. Should the poverty Hue be based entirely on nutritional re- The pe One quirem_ ts, there are obvious implications for the welfare indicator that is used. To be consis Much of the literatre on poverty has been tent with a nutritionallylbased povert line, welconcerned with the resective merits of absolute fare must be measured ondy in terms of food and relative concpts of poverty7 Recent work consumption. As a general rule, we would not has questioned the usefulness of absolute pov- reconunend such calculations for SDA poverty erty, since what constitutes poverty in one soci- profiles, unless policy priorities in the country ety (at one point of time) may notbe the same for concned require it. another society (or the same society at a diffeent 3.23 lte use of an absolute poverty Hne (with tme). Whflst Sen (1983) restates the "absolute its attendant value judgements) canbe better juscase, heretainsmuch relativity. He distirguishes tfffed in situations where there is already a genbetween the "capabilities" which incomes confer eral ageemaent among planners and policy makonindividuals,andtheg0odsandservicesneeded ers about it Of course, the analyst should be to produce them. Poverty, according to this view, encouragd to use alternative poverty lines, and is the absolute absence of certain critical capabili- make comparisons with resldts deived f rom the ties - such as avoiding shame, community par- offiial line. In such cases, SDA analysis shold ticipation and self respect. But the bundle of use the results of the IS to cross check on the goods required to provide these capabilties var- realism of the officia poverty line. For example, ies from place to place and from time to time, and if planners are using a poverty line which was it is in this respect that poverty is relative. If the originally computed some years previously, but absolute aspect of poverty is ignored, as Sen which has been updated using (say) the con- (1983:156) puts it, "poverty cannot - simply sumerpriceindex,itwouldbeimportanttocheck cannot- be eliminated and an anti-poverty pro- whether the inflated poerty line is consistent in gramne can never realy be quite successfe. real terms with the poverty line as originally de- 321 The debate on whether absolute or rela- fined. The conmer pice index may not be a tive concepts are appropriate is simply about the eliale inflator to trace changes in the prices of choice of poverty line, and this is certain to be a the basket of commodities which orignally desubject that needs further clarification at each termined the povy line. country level Ibe anlyst must decide at the 324 The aniuites of an absolute poverty outset whetwer the poverty line is determined in line, its subjectivity and arbitrariness, have led some absolute sense (either b-y policy makers many analysts to adopt an explicitly auitraiy themselves or based on other criteria, such as relate povert line In some cases, poverty is minimum nutrition levels), or in relation to the defied as that income below which a cerain Inome distibution in genral. perei of the population is tobe found. Thus, 47

56 the poor maybe defined as the poorest (say) 40% 3.27 To summariae, one of three broad apof the popat. 8 The poverty line is that in- praces dould be nsideried in calculating the come below which 40% of the population are poverty line distributed at the ime of fte survey. Kanbur ^ A reatt qroach. In cases where thwre is (1988a), for example, selects a poverty lne which no univessally accpted notion of the poverty defines 30% of the population as being poor, and line, the analysis should take the poverty Ine to an alternative bard-core' poverty ihne wldch bersomerelationtomeanlnrme(say twothirds places 10% of the population below it This nakes or one half of mean income). Given its relativist the selection of the poverty line awbiy, but tiothlispovertylinemayvayovertime, tiere is always an element of arbitrariness and depending on whether nm incomes rise or al. subjectivity in deriving a poverty lne, and such a Thisprocedureylelds measues of poverj which procedure makes this quite explicit have no 'absolutist' significnce. Here, the sig- 325 However,thereareatleasttwodrawbacks nificane les not so much in the absolute numto this approach. First the interpretation of the bers in poverty, but in the pattents of poverty poverty line becomes extremly difficult -exactly across the various groups in society. what level of deprivation does a poverty line so * An ablutist aprach. in counties where a generated imply? Are people suffering malnutri- partcula povery line is well understood and tion below this line? Just what level of suffering generally accepted as dividing the population is suggested? There is notg in the procedure into the poor and the non poor, analysts s d used to derive the line to provide any answers to be ecouaged to adopt it. his is patcularly these questions. Some additional assesset, important if the threshold is accepted and used therefore, may be needed to give the analyst and by planners and policy maers, since the findpolicy maker some impression of the extent of ings of the analysis would gain in policy signfideprivation implied by the poverty line derived canoe The priniple should be establshed that in this manner. A secod drawback to this method where a povwey line is curently in use among is that it implies that poverty will always be with planners, it should be utilized. The key point to us - if the poorest 30% of the population are by note here is that in such cases, the absolute level definition poor, poverty will simply never be of poverty will have somte meanin - polky eradicated. And this was one of Sens basic ob- makers and planners wil be interested in just jections to an unbounded relativist approach to how many households are below this level. This poverty. Kanbur (1988a) avoids this pitfall by raises serious questions about whethe the data idenfying the poverty line as that which gives generated by the curent survey designs used in 30% of the population poor in the base year, and SDA will provide accurate esdmates of povrt then reis ts poverty hne in an absolute sense in this absolute sense (This is discussed further for the analysis of later year In the annex.) 3.26 Fortunately, there isan altematve relativ- * A pragmtic apprwc Here, the poverty line ist approaci to the selection of the poverty ne is selected ih an arbitrary maunner for any one which does not suffer from these drawbacks (nor year (using a percentile cut-off, or takin some from the drawbacks of the absolutist procedure). ratio of mean income), and this line is retned in This defines the poverty line ireton to me real tms thoughout the analysis, induding the income (or possibly median income). Thus, for analysisoflateryears Aswlthteflrstapproach example, the povey line ndght be defined as a the absolute level of poverty will not have signifihalf of mean income -those individuals earnin cance -but rather the pattms of poverty across belowhalf averageincomeareconskieredaspoor. groups The extent of deprivation implied by such a benhmarkismorereadilyappreciatedthan with 7eer binda the percentile approac, since the analyst and policy makerwlllbavesomeg apwofthestasdard 328 Apart from the selection of the povty of living to be obtaned from the mean income, line itself, the degr of povery wil depend on and therefore fom some specified proportion of three basic factosm it. Moreover, defing poverty in this way does * the inidee of poverty, as measured by the not imply that povey is ever eradicated. It is numbers in the total population living below the quite possible for incomes to be entirely dit- povrty line; uted above one half of mean inme. * the inniwy of poverty, reflected in the ex- 48

57 tent to which the Incomes of the poor lie below whee e inome gap ra () is given by, thpovertyline, the degee of inequty among the poor, in I.o (1/q)i[ (Y,-Y 4 )/ Y I that htansfing Inoe fno the poowrest to the I.' better-off poor should raise RIwSed "Poverty. It is simply the averag of the poverty gaps ex- 329 Any index or measure of povet shod premed asa fracdtonfthepovertyline. P. orhi ideally reflect all tee of thse dinsins therefore Wmeintoo accunthowpooronaveage Moreover, for our purposes, we need an indhex t poor are, and refets both the indene of which can be used to asse9s the effects of adjust- poverty (as reflcted in H) and its intensity (as nesince adjusbtmnt frequently entals chang- given by 1). It also measures the amount of ining the sectoral comoition of output - from come, under perfect t that needs to be nontraded to traded gds, from Impt compet- traferred td the poor in oader to e y eradiingtoexortngsecs,andfavoringagrcltute cate poverty. Howver, the P,,, measu is in- - our poverty index must be deomposable sestive to Inoame distribution among te poor. across sectors ( br, Transferin income fom the poorest unit to a 330 A useful index which meewts tids requ richer (but stl poor) uwdt wi leave P., unmentissuggestedbypfslgr,gieerandtorbed chaned (as both H and I will be unaffectd). For (1984). Their lam of poverty index takus the tlids to be reflected in the index, greater weit folowing fom has to be given td the poorest Income enng e=(1/nd[(y units. Ttis can be acheved in this cass of pov- P (1/ 1 (Yp-Y,)/ Ypl (5) ertyindicesbyassundrigvaluesof ain excessof I" unity. With a > 1, a transfer of one dollar from where Y denotes the poverty Me, Y, th in- the poorest urdts to other (betw off) poor units come/exwpedture of the ith poor person, n the wl Inrease the poverty in In short, the P. total popuation and q te mnuber of Inome hidces sugpsted by Foster, et al permit the user earners Delow the poverty lin. Essntilly, the to specify a nd threby seect an Index which Index takes the poverty gap of each poor peron rflects his or her aversion to poverty. as a fractin of the poverty line (Y;-Y)/Y, raises 333 The P, dass of porty ixices is flexibl it to a power a, and suns over poor unis Not alsoiin thatitisub lgoup dem (Kanbur, only does this index take into accot the ind- 1987). The "overall" index of poverty can be denceandthe intentofpoverty,tisalsose- shown to cmrs the summation of poverty tive to the degree of Ihquaity among the poor. indioes among al the sub-grps in the popula- 331 This dass of poverty nmswes is flexible ion. If the study population consstsofm groups in two Imporat respects. First, a is a polc or secors, then, paameter that can be varied to rlct rorrectl poverty "aversion. If a -0, it can be readily P. 6mT'sPo (6) shown that (5) dmply becomes, '4 where P. is the poverty index of group j and P.= q/n *H (Sa) thepopuation welghtof group j q.. 4m), TI = 1. As we shal dcov, this d ion where H is the head-count ratio, ie, the propor- propety w e usefu in analyzing poverty tion of total Imrecetving units beow the chngs in the SDA project, since it is possible to poverty lns. Note, if a = 0, it simply meas that generate both overall indices in eah country, the mse is entirey indiffent to how poor and indicesforeachof the region and sodoeooeach poor unit is - it does not matter how far noic groupins under siatin. below the poverty line each poor person is. Thefore, with a = 0, the index is shply the Poverty and.djusbment heldoni rato. 332Alternatively, with a- 1, the poverty index 334 We now cometo the key part of tids plan becmes, nmely the analysis of the inteaons between the adjustment program on the one hand and P.(1=//n)1[(Y.-Y 1 )/ YIpMl (-b) povertyontheofier. anmotinsanmces,data wll not be available on dhanges in poverty- a the 49

58 analyst wil have initaly are data across bouse * 0whuneabkegupwhoremainabove t holds at one point in time as given by the ISW poverty line but are severely affected by adjust- Frm these data, and from the analysis of the ment and thefmerit policonsideo macmos erction ouhle in Chaptr Two, 337 The poverty profile is aimed patcularly links have to be estabhed betwem adjustmt at the cronic and new poor groups. t does not (asfthasntakeplaceoertheyearspreoedingthe attep to Oover non-poor groups who happen suey) and poverty, as evidenced at the time of to be partiar vulnerable to extemal shocks the survey. But before coming to the analysis and adjustmt With survey resudts available proper somethoughtneedstdbegiventhepecu- for only one year, it is difficult to distnguish liar problems raised in having a data set (itialdy betwee the duonic poor and th new poor, and at least) for ondy one point intime. to ensue that all the povert effects of adjustment are covered in the analysis. For example, 7he poor and thewulnrble suppose that the SDA IS survey is conducted shortly after an adjustment program. To what 335 In deciding upon the broad domains of extent can the poverty profie give an asessment policy concern for the SDA project, two criteria of the poverty effets of adjustmet? There are are used (Word Bank, 1990): poerty and m!- three Vpoce goups which are of interst the nerebility. A household is poor if its income (or chronic poor who remain poor after the adjusttotal expenditure) fals short of the stndard that ment (some may have benefited from the poicy society sets - the poverty ine. On the other reforms, but not enough to take them out of povhand, a household is vulnerable if it is particu- erty); the poor who benefit from adjstmet, buit lady open to adverse extemal events or shocks, benefit sufficiently from the reforms to take them and cannot make the necesay adjustments to out of poverty, and the new poor, who become protect its standard of living. Whileit is true that pooras a resultof theadjustment. The IS will fail ultra poor households are certain to be vulner- to cover the second of these groups, and so will able because of their poverty, these are two quite present a somewhat pessimitic picture of adjustdistinct dimensio of need. Some households ment - it will be subject to a selectivity bias, may be poor and not vuleabe, either because giving greater weight to those who lose, and less they are not affected by exten events (as for weight to those who gain, from the adjustmt example, in the case of subsistc famners), or Cearly, much will depend on the timng of the because they can readily cope with the changes survey. If the survey is conducted before the (for example, production and consumption adpstment program, its main purpose wil be to switching in the light of relative price move- fodrecast the likly effects of adjustment - both ments). Others can be vunerable but not poor, a hmful and beneficial -rather than make an ex case illustrated by retrenched public-sector post t. In this way, It muy give a more workers baled asesmd of the gainers and losers 336 Thereare threebroad groupsofconento from the reforms. the SDA Pogm: e once-off nature of the IS, terefore, The chronic poor whoseituation is caused by limits what can be gained from the analysis. It multiple deprivations, suh as low productivity nwstberecognized that the povertyprofile,based due to poor health and nutrition, poor accs to on a suvey following an adjustent progm, productive assets, etc. This poverty is deep- wil be sa t to selectivity bias, and might prrooted, emsting before the recent deterion in enta more pessimistic view of the poverty effects economic r and the implementation than isin fact thecas Even if we are prepared to of adjustent programs It includes the ultra forego coverage of the groups wl" have ceased poor or destitute. Someare vuleale to reces- tobe poor because of the adstmnent, we ae still sion and adjustmen-related shcks, which may left with the dfficlty ofdi betwen have creased tidr poverty fther. Othes in the chronc ad the new poor. Tee may be thlsgwp,however, mayberdatveyunaffected, cases which canbedefined with some confiderne while yet others may in fact benefit fim adjust- forexample,poor snmiboldersfuraldless4are ment lky to be chrondc poor, whea poor civil ser- * he w poor who are above the poverty line vants - or former civil servant - are likely to prior to the shock and adjustment mesur, but be new poor. But there are other poor gpups have faln intpoverty as a result. identified by the survey who uay or may not 50

59 have been poor befor the adfusbmen 3.42 Wealth, wome or expenditures. Wealth is a fundamental fctor affectg houweold behav- An ovffal asxes ior. In nual areas access to land is crtical. - dless or near-ldmes househols can be affected 3.39 U it is assumed that the macro-ameso links quite ienfy from the smalder by adjustare reasonably understoo, and that the real- ment policy. Among the relatively poor, those economy effects of policy can be approdimted with some asset ndiht be distnguishd from (in tems of dhanges in the growth of real output those who have noe. Weal, as with Income or per capita, relative price chan!ges, fraszctura expenditure, has the advantage of being a housechanges and differences in sectoral growth rates), hold-level criterkin However, the use of inoome, the chalege of the data analysis now becomes expenditure, and to someexdent wealt, ascdassione of linkring what is observed at the household fiers, suffers fom a major drawback, in that aclevel with these sectora changes. This exercs cording to any of these criteria, a household's wmilnvolvefourbasicstesintheanalysls First, relative position, and hence its classification, socioecononic groups will need to be identified might change over time or as a resudt of policy into which all households are categoized. Tis invention. For insunce, the mobility of housemay also involve regiona categorization. Sec- holds between income dedles makes total inondly, poverty indices should be computed for come a poor classifier for taetng polc on these regional and sooeconomic groupings. particular households. The wealth criterion is Thirdly, the analysis should establish the main more effective because households are relatively primary income surces (by sector) of each of the less mobile between wealth groups in the short groupings. Finally, an attempt is then made to ormediumrun. assess how each of the categories has been af SodologiaL These criteria include a range fected by adjustment policy. Each of these stages of factors such as race, relgion or language, and is discussed in tum assume dgnificance in African societies where market fragmentation or even ethnc discriwdmina Snml: vw sam ucv no wsccr. The tion might be a common characteistic of the classification of households by socioeconomic majority of the poor. group constitutes an essential element through locatlon Loation is usually justified on out the SDA analysis progr, but especially in the gromuds that policy often has a locational this poverty profile. Somecondderationisgiven to the ies raised in seecting socioeconomic elemt Rura households need to be distin- guishedfromurbanhouseoldsbut,evenbeyond classifiations in the companion volume, World Bank (1990). Some of these considerations are this, there is a strong rpatial dimension in the way policy effects are trsmitted through marrepeated here in brief. Any number of criteria might be used to establish the classification of kets and inf hostruture. Thus it might be impor- tant to use an even finer locational division for socioeconomic groups,butthebasicrequirements the purpose of dasdfying household groups and of a useful criterion is that it shud be unambi- to capture the regional effectsdlrectly. However, guous and have a dear policy focus. Apart from admnidstrativeboundarlesrarelymakeanalyica the obvious importance of policy targeftn& clas- sense. sification should be selected in such a way that 3.45 Oawcrsi of the housd head. The households within them are reasonably homoge- sc of the househd neous in the ways they are affected by (and re- head (e.g. occupation or employment status) are spond to) adjustment policy. Since households often used as criteria for cdlafying households. are most often multi-individual units, a dcassifi- In doing so one is implicitly assuming that the cation should ideauy be dhose which is appli- behavior and level of well-being of all individucable to all individuals in a given household or, als in the household can be determined or adealtenatively, to the household as a whole; othef- quately rprsted by the staus of the head. wise, the fundaent notion of the household However, the economic status of the household being a 'unit' is lost. 3AI1 A broad set of criteria have been used for might be deermined by the charac s of the main earner who could be a different individual dlassify households,includingwealth,nome from the household head. So this criteion has to or expenditure (economi criteria); sociologica; be used with much care in its paccl applialocation; and of household head*" tion 51

60 Box Si.. Assessinig the grouping Ihe key to th seleci of io groups sair oveumesftweois1themeee inomme of gfoup sand some wuwe of _o ywitn the grou In tis PWAU WlItw tcf esd up(oalligthat " = way, households withln a poup can be asmuned to be 1N affecti by adjustment (and to rajond to it) in a dsmilar Theouiveofthepungeinnrdsewoudbetoauxlway. Todwwethv _ he selected4 e Is mi IAehse eand tn o te dgt- f, toihat d, the analyst may wish town the extt to wblh it asmudioftheoveravarancelnln /exp ue epins how bnm are wes ibuted a houhol Inaone vauae, of course, does not prove nomognet e db thegutaseedv If thebetweengrwp vac lains malyamswa proportoof overal wfithn te groupkng but ft can be Intrpreted as a sd vaiane (ay of the cie of only 20% or so), the analyd ndtor of IL A dmple mthod for ddts tat woudd be to de_opos the total vaia in nme (or exedtr) must nspect the wldo_u varac for each group, seddig to idefy grp In whlch bicome vnce i into two broad cnmpor;t - the vaince in trxmnes hewm and IlNkh groups. The grate the proportion of bgh(adwbldicomtrlbute toovemlflvar fbneame/eidltuevarancewllnagrouplsfoad overalowvanexp aineybsyb pwoe-ovaldawn, th mere socrsfulb the Srouphng to onptg to exp a ds*nleaprcpo of he overa varia (ascatpmdwlthtecent ioof omersectors),the incomevaia TheonuMo (v 2 )lnluwnmean isawtthegouplnctse be deamposed aslb:ows ous for anlyl purpoes The anayst may cosder V2+ ej+j(pj subdiig the poup fuder to order enha to bhome- _p)2 I7) SwZ1geeltyObviouslythemre aredfsngw9ah 1-61 ~ 1*1 ~ ~ th do utwdexhe e y y power of the 8rub DAe ithn behtee dedion of wberi to stop furt he and live roup goup wdth thep,pordmo of vaianceelnedby thegroupig vaae vaance bas In the lt analysis to be a dgement for the analyst whor ol, is hey vrnce Ininme whifif,oupj4jis the hfs 3.46 In thecase of Sub-Sahan Africa, an indi- ment If, for example, macrwo-meso analysis discation of some broad categories of sooeconomic n betwe a number sectors, it would be groups that might figue in a number of txono- usefl for the seection of no categomres is as foliows: ies to uilize this setorel dhl How- Rural saetor ever, this may lead to a serious problem for the * export-oriented medium and labe fames analrsis of IS daa - the pblem of empty cells. * export-oriented smaihole Too detaed a d _PSaggngton for soi-eonomidc * food/s /bstee farme groups m t lead to small numbers appearing * poraists in the given hr the reatvely small sample usu- * laxless(orn rladless)agrulalworkers ally ontemplated for the IS * no griltura worers 3A8 In tde last analys, th selection of so- Urban sectr economic categories must be country specific. It * government empoye (skilled) mnst not ordy reflec the specific rha tst * government eployes (unskdlled) of the cunary, but the poicy priorities of the * private/forma sector employe govm n It is important that policy makers *private/fforal sector employes are fully conslted before the socoeconomic * infom Sector self-employed lfcatonof hous olds finazed. he pur- * inacdve or unemployed pooe of the cassification is two-fold: it nakes the 3A7 The above claificaton may be fuer data analyis more nageable; and secn it rdined by betweenthecapitalcity should enhance the ability of governmnts to and oter urbanarea he prcy of the cspital design poldes whih can assist e gmups in a city, which often makes it signifiantly dierent tarseed way. Because thepoorestxperaenlein from other smll urban centers, sts hathis most developing countries of Afica is het distnction is important in data analysis. If pos- gous, inome distution data per se are gensible, the selection of soleconomic groups erally unhephl for practical policy dedsn. should bear some relation to the structural Cingfteanalysintersofradlydfiae changs that have been brought about by adjust- Ioeconmc grups makes it more policy rele- 52

61 Frame 3.1: D mso of F poeymesum by socup, COte 4 Jvole, 1985 (30% cut off t P,Qo P,,= P,c2 Coarawo CO*&%"CORa#tb Sod8eamm.kg~uap Value (rtu Vae C wa) Vlu (jwr) Export femurs Q Q Q080 1&8 Food pw s Q Q184 6" omlem9oymr,wmm Q0. 13 Q.IIQ Q2 0.00w 0.1 Poemployeaw pdv*a D Juformals mto A 0Om 14.5 O A nna psut Mwgvtg% t *popue as poor. vant. Box 3.1 outes how a simple analbs of food-cop fmem and ordy 3% of goverunmt variance can pwvide some insight hnto the use- emplyes were poor. Analysts shold not that fulness (or otherwise) of the socioeconomic the nand q (and of course Y, values are different grouping selected. for each group. So in the calcation of group povety indexes, q is the nunber of poor in the 3A9 S9u 2: P 6 CATboNS. Having decided group, and n s the population in the group. upon the regional and s c dafi To decowmpose the total poverty index tions of households, the analysis proper may be- across the sectors, each sectoral poverty index is gin. We suggest at the very minimum that con- multiphed by its populatioki weight (), and this putatlons are made of the P4 indices (based on value is pessed asa proportion of the poverty tot household expenditure per capita or per Indx for the population as a whole. Thus, for adult equivalent) for a =0, 1, 2 (or 15). These example, for export farmers, the computations shoud be comnuted for the coutry as a whole, are as follows: for the regions of the county, and separately for eah of the catories. Any esti- (P..d)( Z/) f Po. - contribution mate based on els In which the number of obsevatons is less than 30 shodd be explicitly (365)(.186)/.30 = 223 (or 22.3%) identified. Estimats shoud be reported for a number of overty ies (for example, two lines Thepoutionofeport famerscomprises18.6% representing one half and two twrdsof mean per of the total population of CMte d'lvoire, so that capita Income, or gfng 30% and 10% of the with P. for eport rmkers being 36.5% and for population as poor). Frame 3.1 provides an ex- the contry as a whole, 30%, the contribution of ample of thebasic typeof abulation which soud exportfamerspoverttoaggeteperty(with be prepared under the poverty profile, taking a a = 0) is 223%. For a = 0, the interpretation of percentile approachtodefinngthepoverty.i" this Is very straightforward. Itsimplymeans that 3.50 Before we discus e inferences that can 22.3% of the poor are found among export farmbedrawnfromsuchflndings,ltmaybehelpfulto em. Decompos poverty across the various establish exacdy what the figures reported in groups (given in the second colum under P 6 =O) Frame 3.1 mean Becmuse al the P,mnasur take tells us, for example, that only 1.9% of the total the poverty gap as a proportion of the poverty poor in the country were formal-sector employline, they ae all in the range of 0 and 1. As a ees, whilst 59% were food producers. Inreases, P, becomes nmaller, so that P,-, > Pe The ntpratlon of P., is a little more Beginning with the simplest case where a = 0, the comphcated, but neverthele an intuitive expla. poverty index is simply the headcdunt ratio (H). nation is possible. With a = 1, P, equals, where In 1985,P. wasequalto0.30forcoed'1voireas I is the ince gap ratio. Recallin equation Sb a whole, which means that 30% of the total pop- above, * isclear that I s also somewhere between lation of the country was poor. Similaly (reading 0 and 1. It is the poverty gap as a proportion of down the first coumn under P,.), nearly 50% of the poverty lne, aveaged across all poor units. 53

62 Since both H and I are in th rae 0 to 1, P,, eradicate poverty among thei coleagues. As (which = H) is also in this range. It also fiollow withp. the W Indexcanbedecomposedacmmss that H > Hl, or Pr > P,. Thus, for example, sector Ths shows how fe aggregate per capita reading Fme 1, we con derive the inwome gap poverty gap is distributed across the groups. Of ratio (1) simply by dividing P. 9, by P,> (or H). the resoures needed to erdiate poverty (*his Thus, for the ountry as a whole, since P.,.10, being 10.3% of the poverty line in per capita it follows thati is given by tms), 64.1% would go to food produces, oldy 02% to govemnment enployees, and so on. I = HI/H = P,,,/Pe0 =.10/.30 a With a > 1, no such straight forward intuitive interpretation is available, since the index 3.53 It remains now to explain intuitively what now gives greater weight to the poorest groups. I =.33 and HI-.10 really rwan. Beginning with However, the principles are the same. The first I, this simply mean that the gap between the coltumn under P.2 gives the index computed for poverty line and the average income of the poor each sector seprately, and the seodnd column is 33% (or one third) of the poverty lne. Sine, reports the decomposition of the aggregate index across the groups. As with the other indexes, the (Y,/YV) I - (Y- Yp up indexes are computed for group-specfic values of n, q and Yi, and the decompoition it follows that the average income of the poor is applies population weights to the group values two thirds (given by ) of the poverty to genrate contributions to overall poverty. We line. Values of I for each of the groups nuy also tum now to review some of the implications of be computed and interpreted in this way. Thus, these findings. for example, the income gap ratio for food pro Frame 3.1 shows the usefulness of the deducers was (a 0184/0.495) and for govern- composition propert of PAin underst how ment employees it was.065(a 002/0.031). This poverty isaffectedby adjusent. Assumingthat means that on average, the incomes of food pro- it is known how the various groups are affected ducers are 62.8% (=1-0372) of the poverty line, by adjustment (a subject to which we return whist those of govenmuent employees are 93.3% shortly), a great deal of information can be derived ( ) of the poverty line. This highlights the from a simple table as this. Taking the incidenc extrainfonationthatisprovidedwitha=1. We of poverty first (o = 0 in Frame 3.1), dealy the not only know that the incidence of poverty is poor are mainly to be found among food progreater among food producers than say govern- ducers (a sector containing 59% of the poor in ment employees (through the head count ratio), Cote d'ivoire). 22.3% of the poor are export but we also know that the intensity of their pov- famemr and 15.5% work in the urban informal erty is greater. The average poor food producer sector (or are unemployed). Peverty is hardly eams an income of ondy 63% of the poverty line, present among formal-sector employees. Notiee whilst a poor government employee earns on that the ontnbutiontopovertyof foodproducers average just over 93% of the poverty line. rises (from 59% to 65.9%) as higher values of a 3.54 Recalling equation5babove,itiscearthat are taken. Higher values of a mean that greater P., or HI sums the gaps between each poor weight is being given to the poorest groups. This person's income and the poverty line, and di- suggests thatarelativelylarge proportionof food vides by the totl population. It is someimes producers are among the poorest of the poor. If refered to as the 'per capita aggregate poverty greaterweightistobegiventhoseintheextnes gap'. This gives a measure of the amount of of poverty (thatis,takingta-2) almos two thirds income in per capita terms that is necessary (un- of nmasured poverty in C5te d'ivoire in 1985 derperfecttag toexadyeradicatepoverty, emanates from food producer It shows that if every meber of the population 3.57 Frame 3.1 can readily be extended to inchipped in 103% of the poverty line, there would dude regions of the country. Tabulations which be just enough to bring all poor people to the reportpoverty bybothsocioecnondicgroupand poverty line. Similarly, if all food producers region will be particularly useful to policy makchipped in 18.4% of the poverty line, there would e. In the case of COte d'ivoire, for example, It is be enough to bring all food producers up to the essential to establish where the poor export-crop povertyline. Govenmentemplloyeeswould only producers are to be found. A tabulation such a need to contribute on average 02% to exactly Frame 32 can readily provide this inforation. 5'

63 Fame 3.2: Incddence of poverty by region lins. 3This test involvescoumaring the cumulaand o c group: COte dvolire, 1985 tive distributi of inome/expendite up to a30 cut Off) some maximum poverty line of all pairs of sodof(vvmdv economiccategariesthatare isd u A --- Wea -O&n- E-tst am- if the cumuladve income distribution of one w AW dm u w& t S n caoeom idc group (say group A) Ues nowhe *~wimmo.ukgroup d-n tust, jb'esi faust nih AU below that of anor (group B), then povertycan Bxpat.a" famas be said to be unambiguously geater in goup A Food-rop arrs than group B. This is called the first-order Fo_mal nmt dondnance test. 14 It mea that for all admissible FO.5LP*t poverty lines (that is, for all lines below the Infomal AU QS stpulated maximum perty line), povet in group A wil be greater than among grup B Sowec Kaab'ws regardless of the choice of poverty measre. s This domdnance test would be applied to all pain Most of the poor eport-crop producers are lo- of sodoeconoc gpups. In most cases, the test cated in the Savannah region of the country, and should prove condusive, and the orderig obthese are maindy cotton producers Similarly, the tained with the P index can be considerd rotable shows that the majority of poor food-crop bust But where the test is not condlusive, the P. producers and informal-sector workers are to be orderings should be suitably qualified. uind in the Savannah region. Such two-way tabulations can be extrmely helpful to policy 3.59 SIm 3: meaufic PRI my mcme sou maker and can be computed for other values of The analysis thus far is based on fairly aggregaa(frame32iscomputed fora = 0). tiveclassiflcationsof households. Insofarasftis 3.58 Therankorderingofs n groups based on hwrme sources, t relies only on the according to their poverty status is the key to main occupation of the household head or on the most policy analysis. As we have already em- prindpal crop grown. There are two main probphaszed, where the poverty line is explicitly ar- lems with this type of analysis if some judgebitray, it is the pattern of povrty acros groups ments are to be made about how adjustment is in sodety that is Important In the above ex- likely to affect the poor. First, a greater degree of anple from C6te d'ivoire, the poverty ordeings disaggegation is usualy required for policy remain the same, regardless of the value of a that analysis. For example, 'export-crop producers is chosen. '2 But this need not always be the case. should be further disagegated into cocoa, cof- It is possible that changi the poverty index, or fee and cotton producers, since adustment polchanging the poverty line, will yield different icy is liy to affect each differently. But such rank orderings. To what extent can we be confi- disaggregationsoften lead to small ce problems, dent that the poverty orderings obtained with given the relatively small sample size of the IS. the P, class of indices would also apply if another Second, the sooeconomic classification can be poverty line were chosmr, or if another Lidex of based only on the main occupation of the housepoverty were adopted? An assesment may be hold head or the main crops grown. It only acts made of this through the use dminance condi- as a rough guide to the income source of the Frame 3.3: Sources of income by socoeconomic group Peentage of tete bsby iameuoui Agrkd*ue No*.f SocArwnausgimp We enepise ex eis. Redt Tnaffs Export-acrp farmers 100 Food-cop frm. 100 PFormL government IOD Formal privae 100 hiformal 100 An

64 frame Mure of poor household pumuy Iume by gc p and secto of orig fppmdw) ~ ~ ~ ~ ~ TvdleNn*dk Cant- GAO!ww- Agriadahv MW&ndWtuh ~frct mess Oilie Eupoit Feed 1p2rt- Pi'tetdr Sioac&wm*nkvUP Civ, crop aie Imiorbk Other Total Food-acrphfarmTrs 100 Formail goverment 1OD Formaul private 1_ hafornal 10 Box 32 _wome sources: An i1ustration household conred, and during tmes of adustment, it may be a very weak guide indeed. Asan lustration of ate "pe of tabuaton sggerted This b for two teasons. First, the houseld head by Fw33Sand3A, the fofowingiebsed on te maynhwslrlpyngeinmultipleo=cupati reslts of Ome U sure In Cke dlvoire 198. (or diversify output) in order to cope with dete- S sof mee Inome tervad rgazioupk C8W ke, econoioraticg cns or hmr w ot main ocupadon may be one of a number. Sec- (_ff="4 ond, other household members are usuaiy engaged in differet occupations (or grow oher Ha.whol*.aops), and may in ha contribute more to tot AO Fam houseold hicome thandoes thehousdwhdohead. osuoid hosscad& For these reasons, it is necessary to bace in more Imewr AU Lare s=l detail the Incme sources of poor households I tion on sources of income Is ob- Coffe tuned in the IS, icuding farm and non-farm Oferop enterprs Inome, wage income ard other in- Home produced food clme soue (such as rental inow and isnome Uvetock A transfers). Fomnthesecomputations,tabulations other ruhoomue QA 0.S can readily be prepared to trace the main soures rota tn oe) OA O (4) (40 of Inome of poor households. Frame WageInme the type of tabulation that would be requlid for Rets, dkndsterest e 38U this. From such a table, it would be posble to Not homier innome check whether the Income source of the house- Net business me 19.8 ItO hold head (or the main crop produced) is a useful Commsumdurabesenwls indicator of the main source of ince of the sd = nco.) (69.2O ) OX (55 household. Thus, for exmple, we can assess Total 100 1W0 1OW 1OQ whether a household in which te head is em- MnmonNmlyincome ployed (say) in the pic sector lo receives (CFA Ft. UQ 1S8Q4114S most of its income from that sector-or whether NoW am*boftn%wwswm6ema=v4_n9fhw=,whwh the household (both the head through other occupatis and other household members) eanrs Nota w_ i signiw tficat income from other activitie. This is a rwbedinesnasmurniwagg_oupu critical from a policy perspective. It would be quitemiseadingtodrawthecondusionthatpoor fhaodho inmo households whose head works for the goverbouodwd _wionm.ment will lose out from public-sector wage km j_n. v CM freezes, if that household is also engagd in ac-

65 tivitles whidch are liely to become more profit- Frame 3: Ace to edt byoeconoic able (say wage incone from employment in ex- group port mn fa. Siilarly, it would be es- (y,fwstobmd*xs oam o pvywotpmxaao sential to establish how much other income huw& muiibk IOEtMb 0 Oww)" soures are tapped by agricultural produce rr Frame 33 hould be computed for both poor and S n_ rgw &.ag* m AU non-poor households separately, since income sources among the poor may differ from those of EXPOC aopfann the non-poor within each o nmic group. Fod-pfamas 3.61 These data are helpfl in identifying the 0 diversity of inconme sources of the household. a However, we need to establish how te different Al sodoeconomic groups are affected by adjustmt It would be helpfd, therefore, to trace the sector exportables rather than importables, and nonof orgin of income eamed by the various groups. proteced inportables rather than protected im- For this, we need ordy be concned with pri- portables? If the poor are to be found mainly maly income- that is, incomeeamed from pro- among food produces, for example, reading ductive work. Frane 3A presents the type of acrosstheappropriaterowsinfranmes3.3and3a tabulation which is required for this. The disag- wll reveal whether this group is engaged in aereation of the productive sectors is designed to tivities in which retums are likely to increa as a separate out tradable from nontradable, export- result of policy refoms It may be that a nonable from importable, cted importable from trivial proportion of the incomes of food-producother Importable, and government from private ers is gained from wage income (as evidenced in rkontradable. These disaggregations conrespond Frame 3.3), which means that such groups mdght to the logic of the SDA cocptual framework be influenced in two broad ways by the adjustment. They are affected as food producersby the 3.62 Siw 4: AN Assr j of mn EFCI3 s of effects of adjustment on food markets (which ADusniEr. Having these basic tables prepared, may, or may not be favorable, depending on the analysis must now attempt to interpret them, whether food is a tradable conmodity), and also and to make some judgement about how the as suppliers of labor. Tlhe likey effect on the poverty groups so identified have been affected latter will depend on the direction of change in by adjustmnt. This analysis should proceed in the real wage and on employment opportunities two stage The first is conducted at the group (again, see Chapter Four). level, and seeks to identify how each of the sodo- 364 This analysis provides ordy a static pue economic groups has been affected by adjust- of the implications of adjustment for poverty in mnt. he second delves into the issue of within- the country. It is also important to make some grup differences, concentrating on those gups judgment about how poor households are likely in which most of the poor are found (as evi- to respond to the costs and opportunities associdenced by Fame 3.1 above). ated with policy refors. Some households may 3.63 Gmuplevd anass. There are two main be producng the types of commodities that are questions that need to be resolved here. favored by adjustment, but may not be able to * Are the poor engaged in economlc activities raise production, or may not have access to marwhich are favored by the adjustment program? ket oppordties. Similarly, households may be * Are the poor able to respond to the incen- produangproductswhkcharenotfavored,sofor tives offered by structural adjustment? them the critica quesdon is whether they can The first deals with the current activities of poor make the neceary adjustmts to move into households. To answer this question, the analyst favored product lines. To get some idea of this at simply has to relate the findings of Frames 3.3 the group level, the analysis should nake some and 3.4 to the results of the macro-meso analysis assessment of the access of these groups to prodiscussed in Chapter Two. At the time of the ductie assets, on the assumption that this wil survey, werethepoorengagedinactivitieswhich reflect the abiity of houseolds to realize the ;re likely to expand as a result of policy reforms? potntial gains offered by adjustent. If house Were poor houwholds deriving their Incomes holds hold key produceive assets (such as promainly from trdables rather than nontradables, ductive land, livestodc, firn equipment, tools, 57

66 hbor etc) and ff they have ready access to in- was somedtng that we deiberately suppressed porant inputs (such as ferilizer, extension ser- in the betweenrwp analysis. For example, asvices, irigatnwater, credit, and so on) It is mome that adjusutent favorably influenes exmoie likely tat thy can respond positively to port farmers as a group, so that other tngs bethe beneficial (and even the unfavorable) effects ing equal, the mean income of this gwup wil of adjusmnt. he IS questionnae obtains in- rise Does thsnecesary imean thatpovertywin fmadon on many of these assets and on the be reduced. This depends on whether withnacoessofhofseholdstd awkets,lnputsand cedit. group variadons In income change at the ame Fe following table gives an imustatlon of the time. If income varianc rmais unchaged (or type that is needed for ths part of the investiga- Is reduced), poverty will fall. But If witin-group tion. Of course, similar abulations should be varce actualy rises as a resut of adjustnent prepared for other assets/inputs. Fram 3.5 il- (for example, through a reduced access of poorer lustates two types of data that are obtained export farms to credit), there can be no prethrough the IS. The first concems the amount of on that poverty will be reduced. Ths is credit actually obtained by the household, whilst why some analysis of the withngroup variation the second is derived from a specfic question in inincome/expend isanecssaycomponent the IS questionnaire on difficulties in obtaining of the pov profile. credit. Whichever approach is adopted, the pur The dependent variable for the multi-vanpose of the tabulation is to uncowver any signifi- ate analysis should be the same as that selected in cant problem encountered by socioeconomic computing the povertyindex-takeninourcase groups In obtaining credit, and to establish to be per capita total household expenditure. We whether poorer households experience greater specify the following gera estimaion equadifficulties (in this case in obtaning credit) than tion for wihn-rup analysis, other households in the same socioeconomic group. If thislthecase, the analysis would Y- Yl A, P, F, (8) suggest the incentives of the adjustment progrm may not be sufficiento raise incomes among the Per capita total household expenditwe (Y) is aspoor (even though non-poor households in the sumed to depend upon five broad groups of resame socioeconomic goup seemlikelytobenefit). gressors household istics (H), asset This would suggest that adjustment policies need holdings (A), a vector of relevant price variables to be complemented with credit-market inter- (P), a vector of factor price variables (F) covering ventions to assist poorer households in gaining labor and credit prices, and a vector of appropicredit. ate infrastructural variables (a), such as access to 3.65 Wtin-grou anysis. Tabulations of the roads, education/health services, credit, and so sort described fius far have two basic linita- on. Equation (8) is a reduced-form of an impliit tions. The first is that they aggregate across strucural model in which Income/expditure households within groups. In most cases they of the household is viewed as the outcome of a describe grup means, and do not give any im- decson-ming pess The household i asw pressionofwheer e are signicantvariations sunied to mandlze utflity and net income, subwithin the group. In effect, the analysis has as- ject to its resource wnstraints This process desumed that withn-grup variadons in comes pends on its characteristics and asset holdings, and other relevant variables do not hange in and on the market opportudties it haes As a response to adjust. This pedtted us to use result of these decisions, there will be significant group means as the basis of our analysis. If you differernces between households in the ways in like, the groupings were selected to reveal the which they genate Incomes Some households majoreffectsof adjustent through theirinfblence possessing productive assets wil rely on family on between-group differences. Secondly, as er Ots, holdingonly theirlabor, will tabulations, they can ordy take into account a depend mainly on wage income. It is these diflimited number of influence fees which formed the basis of the socoeco Theanalysisshould therefoebe takone nomic grouping discussed abovu stage frter, through withigou multi-vari The reduced form, therefore, wil obviate analysis. In this way we can investigate ously depend on the income-enig opporthniwhether adjstment is likely to influence the ties facing households. As a producing unit (op. wihin-oup variations in qexediture, which erating either an agricultural or a non-farm en- 58

67 tepris), a househows income will depend on Frame 3A6 Expenditure pattemsby poverty its productivity, wnich in turn depends on its goup: Ghana, holdings of non-labor assets Itwill aso be deter- (rn8ta) mined by the costs of purchae inputs (such as labor and ferffilizer). On the other hand, house- AU Mmpoo Pat holds relying on wage 'bor will find that meal Maketaexpditure wage rates and their numan capital assets will Ofood have a critical influence on income. It is dear, then, that the specification of (8) wiu be quite Comsptiouof different according to the source of income of the hprducd fd IOD household. Therefore the reduced form shoud Cp*ffpmof is 19 be esimated separately for eah socioeconomic produed 10 84S 15.5 group, since we expect to obtain different coeffi- non-food items cient estimates in each case. It may not be necessay to investigatexpenditure variations within expenir I groups where very few poor households are to be found. From Frame 3.1, for example, there Remlttaepaid IA would seem to be little point in investigating out tf7 withi-grup variations among fonnal government employees, since only 3% of such house- topenditure =ncom is holds were poor, and poverty among them rep- Mdkird resented ordy 02% of total poverty (with cl= 1). Within-group analysis, however, would be fruit- Total expediure ful for export-aop farmers, food producers and informal-sector workers. s oaens et al I 3.69 ThIsexeriseIsintendedtoestablishwhich of fte right hand side variables significantly i- fluences per capita total household expendi- economy. The level of disagg tlon that wil ture. In this way, it will estabush which factors have been feasible for this part of the data analydetermine fte varitions in income/expenditure sis wil be cirumscbed by the detail that is within the group. For example, are export-crop avalable on the nmeso-econi effects of the adhouseholds poor because of their location, smau justment. However, in many cases, adjustment land-holdings, crop mixes, low levels of educa- wil also entail the deliberate manipulation of tion, high dependency rates, use of and access to key prices, through changes in tax and subsidy inputs or credit, access to markets, etc.? Having policy. This may be to achieve an Improvement established which are the most important deter- in resource allocation, by takig a domestic price minantsofincome/expenditurevanation,weare nearer to the border price, or simply to reduce in a position to understand why some house- the fiscal deficit. for whatever reason, the analyholds in the group are poorer than ofters, and sis the poverty effects of an adjustnent program whether such factors are likely to be affected by is likely to be called upon to assess the effects of adjustment. Som of the right hand side vari- specific price changes. ables are unlikely to be influenced by adjustment 3.71 The commodities selected for this rather policy (such as land size or dependency rates), more detailed analysis are inevitably country whereas others (such as market access, credit, specific, and will depend on which prices are and so on) may be signficantly influencd by liely to be mandpulated by the gvermnt and policy refonms. In this way, a more complete for which reason. Te first (and obvious) point picture of the Implications of adjustment for for data analysis is that it must take into acotmt poverty can be ganed. the effects on the poor as both consumrs and produoss of the commodity in question. The Sdective price ievetos household data set geatd by SDA will est*b lish, in the first place, which g ps of house Thewlysisthusfrwillhaveestablished holds produce and conume the onunodity in the poverty effects of the adjstent in terms of question. Without this critical piece of inhformaits gneral r ssons on the stucture of the tion it would be quite impossible to make any 59

68 I I ~~~~~~~~~~~~~~~~~ t3i2i. liiil lq t5['} I it

69 that is requred to avoid poverty. But even the In fact countact the direct conumption effects. poor night be able to exercise some choice In It I possibe to take some account of these feedtheir consumption behavior, nd to Impose the bacus using basicaly the same Vparal-eqil requiemn that a certain fixed burdle of com- uium' approach proposed earlier, by makig asmodities is needed to avoid poverty night be an sumptios about what the liy wage response unneessway vaue judgement for theanyst This ndght be to any eiven price rise. An example of is less dear for the destitute poor, for whom there this can be found in Ravallion (1989).17 are few choices. Hungry people need to be fed, and a case can be made for applying a restricted Basic needs and the poor commoitybundle in defining their poverty. But for other poor groups, some cnwsumption flexi Povert is a multi-faceted condition, and bdity may exist many of Its dimersions will not be propedy cap To correct for this, RavaUion and van de tured through the analysis of household expn- Walle (1988) propose measuring poverty by us- diturea To complete the poverty profle, anainga suitable utiiy Mdax Their method, which is lysts must direct their attention to other aspects based on 'equivalent income!, takes into account of poverty, focussing on the basic needs of poor household prefern in deriving the poverty groups. In the conceptual framework of the SDA measure. In effect, their method ranks house- project (World Bank, 1990) the socia infnastholds by a money-metric measure of utility, with thre is specified as an important element of the the poverty line defined in the same space. 'meso' economic system - lindng poor house- Householdsexerclsechoiceinsletngthebundle holds to macroconomc policy. We consider ofcomnoditiesrequired to achieve whatver level this to be a critical element of adjustmnt/povof utility ispenritted, so that they are notobliged erty interactions. Not oly can access to human to consune any pecfied bundle of goods. 16 They capital enhancing services increase welfare in the awe considerdc poor only if measured utility fals short ran (improved health, for example, being short of the poverty line. We shau not go into desirable in and of itself), but it can have farfuther detail here on this method, but it may be reahing economic effects into the long run - that some ticipating counties would wish to enhang also the productive capacity of poor adopt this approach as a ans of taking into households (thus placing them on upward 'inaccount the consumption responses of house- come escalatorsd). It may be posible to esinate holds these longer-run effects on productivity of a 3.77 However, the responses to policy-induced change in human capital among poor households price changes wil not only be the conmption (for example through enhanng their education response of the households concerned. There will and health status), but this should not be a preocbe denmnd responses by non-poor households as cupation of the poverty profile For present purwell supply repos. These all combine to pro- poses it wiu be sufficient to trace whether their duceproductnmarketoutcomeswhichdifferfrom access to services has been affected by adjustthe price outcomes ass d under first-approxi- ment, and whether there is evidence that this is mationapproachoutined above. Similady, there having noticeable short-run effects on poor may be reprcuaso in other markets which households. The longer-nm effects will just have significantly affect poor households. Take food- to wait unti later. deficit poor households facng a policy-induced 3.79 As with other elements of SDA data rise in the price of food. The above analyis analysis, a tw-stage approach would seem to be would suggest that the households would be caued for here. The first stage should analyze made worse off, and measured poverty would how adjustment policies (notably fiscal adjustrise. But the rise in food prices might induce ment) have affected the awat and qualy of increased food production among (non-poor) services acmss the country. The second stage food-surplus houseolds, and a general increase shld assess whether these changes in the supin the demand for labor and wage levels. If food ply of services have interacted with demand facdefict households benefit from tese improved tors to change outwm at the household level. conditons In the labor matet, the full effect on 3.80 The first stage (the macro-meso stage as their inomes will not be as adverse as that indi, discussed in Chapter Two) will involve tacing catedintakingthefirst-orderapproxtmatl This how adjustment has affected the resources that feed-backresponsethroughthelabormarketmay are available for health, education, and other 61

70 Frame 3.6: Mean number of days riactive due to tneuss Ghan Frate 3.9: Percentage distribution of i1 people by tye of onmultation and povety Ragsm Very PAW' 8D:p Ghana, / Ru.) S.0 S.0 53 Vetry Alr AU Acaa Other urban S Docw Sewc8Ioetenget al 098). NUwe M?v uswnt lj 32 Qo 3.7 lt o0 Other S2 61 6C2 None AN S2S 10D.0 programs of social support (eg nutnition pro- Sm Uootes eta (9low. grams). If it has involved fisal cuts, the analyst must identify how the cuts have be distributed across the various functional components of the fluences of supply- and demand-de fctos In exchequer acconts. By applying appropriate other components of SDA analysis (notably the price indices, I these nominal chages should be analyses of health and education in Chapters ive traslated into real resource adjustments facing and Six, respectively), more detailed analyses of eachofth sectos Thesecalculationsareieeded these relationships is presented. The poverty inordertoplacetheanalysisofthesdadatainto profile should simply identify the level of the the context of the adjusttent progm. They indicators for poor and non-poor groups, and should then be combined with the community- seektounderstandthemninthelightofthechanges level data that the SDA project will generate. induced by the adjustment progam These data will include infornation on the availability of education and health services, and Helth on the quality of the se7ices (covering for example, the availability of drugs, textbooks, etc.) Tabulations on health should achieve two In this way, the analysis should identify how the broad objectives. First, they should identify eal expenditure changes imposed under the dis- whether the incidence of ilness (or injmy) is cipline of adjustment have affected the supply of greater among poverty groups than elsewhere in social-sectr s es in the communities covered the community. A table showing the percentag by the survey. Various indicators can be con- of individuals reporting ill during the referee puted (teacher/pupil ratios, number of doctors/ period specified in the IS, by poor/non-poor, rehospital beds per capita, non-salary recurent gionandsocdoeconomicgroupwouldbeausefud expenditures, etc) which wil give an indication startng point. Unforbtnately, illnesses and injuof how the observed changes in resource availa- ries are self reported in the IS, and ex ence to bility have affeded services across the country. date (with the LSMS surveys in Ghana and COte Initially the community data set wiu be for just d'ivoire) have suggested that better-off houseone point in time, but as the survey is repeated, it holds are more likely to report ilness. This does wil be possible to have more accurate estmates not mean that the incidence of illness is greater of how services have changed at the community among such households. A second broad objeclevel. It might also be posible to distinguish tive is to establish what are the main repercus those regions/communities where the poor are sions of ilnhss, in ternn of the number of days maindy to be found. In this way, an asessment lost through the illness and consultations with may be made of whether the burden of fiscal medical personnel Tabulations on number of adjstments has been bore disproportionately days lost and type of consultation by sodoecoby poorer commumities t nomic group and by povet group would be 3.81 The semond stage examines the meso-mni- conditional on the respondent reporting an Imcro relationships ustng the IS data sel The pov- ness. erty proffle should make some assessment of the 3.83 Thus, tables should be prepared on the health and educaton status of the poor, through i e of illness, numbers of days lost and tabulating a number of key indicators These type of consultation by poverty group and by indicators are the outcomes of the combined in- socioeconomic group. Frames 3. and 3.9 are 62

71 VwyTpos. Paw All.- Frame 3 Literacy and numemcy rates by Frame 3.T1 Paentage of children (aged 12- region and povrty group Ghana, ) cuently attding shool by region and povety group: Ghana, mw Rqio Uis Nit.y Nummwwwj Lere Nta_y UI*NWP Nay poor Anr AS Rurd Rural Aa 4Q7 5u * Awe Other urba * 61A Oh urban20 2M *0bvematW=- Sower BUoats at 1(1969. smwbotq4 at al 09. again drawn fom Ghanaiaa data. The mean sigrfantly worse among the poor, and rurl number of days Inactive due to illness (condi- areas a consistly outperformed by urban artional of course on being ill) is less for the poor eas It is not surpriing to note that both literacy than the non-poor in Ghana. This is to be ex- andnucyarehighestamongthebetteroffin pected - the poor simply cannot afford to lose Accra. Status vrls, suggest that educational working days through illness. Neverthless, the disadvantageislikelytoaffectthenextgeneration fact that on average 5 days were lost over the of individuals livinginpoorhouseholds. Among refeence period (of 28 days) among very poor ill the vety poor rural households, only 40% of persons, represents a serious problem for such chiden of secondary school age are attending people. Frame 3.9 also presents some sobering school. Tbis compares with 58.5% for poor and statistics. It shows that more than two thirds of non-poor combined in Accra. Obviously, such the very poor who are ill do not conlt with any tables should be prepared for other levels of medicalpersonnel(whichcompaesunfavorably schooling, including primawy schooling if unifor the in population as a whole, among whom versal primary education is not yet attained. Noorly a half consult no one). These results should fice in Frme 3.11 that the number of observabe compared with (and possibly explained ffons for some of the cells is very low indeed. through) the macro-meso analysis conducted Such instances should be identified explicitly. under the SDA program Again, the analyst must attempt to relate the findings of the micro outcomes observed for edu- Fda tion cation to the macto-meso anysis described ear Some simple tabulations should also be prepared to highlght the education of individu- The dynamics of poverty als in poor and non-poor households. Again, two simple groups of indicators present themselves Thus far, our concen has been with the First, there are 'outcome indicators, which re- poverty analysis of a data set at a given point in veal the effects of past education exprience and time, assuming that the IS is appled only once. pasthumancapital investnents. insoffespects, However the IS is likely to be repeated after a these indicators can be considered as part of the passage of ftme (say after three or four years). If cause of poverty, espedally as they apply to adult we have observations for two or more points in members of the household. These are mainly time, the opportunities for analysis are widened literacy and numeracy rates. Secondly, tabula- considerably, toincludeananalysisof changesin tbons on curent education 'status' of household poverty over time. These dynamic issues are members shoud be presented. These indicators considered in this section tell us something of the effects of poverty on the 3.87 The timing of the IS surveys in relation to human capital of sucdng generations. Frmes structural adjustment is critil in interpretng 3.10and 3.11 giveexamplesof thetypeof tabula- the data, especially where data for two or more tionsofoutcomeandstatusvariableswhichprove points in time are availabl. An understanding useu of what changes have occured in the economy in 3.85 Notice, the poor and very poor compare the intv between the sureys, and the extent unfavorably in the cas of both outcome and to which these dhanges have been brought about status variables. Uteracy and numeracy rates are by adjustment policy. Much of the over-time lier. 63

72 poverty aunyis wil involve repeatn the tabu- inequality (keeping meanincome constant). Let I- lations that are deribed above, and comparing Pb be te overall povety dex in the base year the two sets of results. he aim of such ompaui- and P be thehinex for the temdnl year. Now sons is mainly to establish wheher the cross- denot asthpovertylev that would have section inferences at are drawn through the occrrednif the cha geinthe in point-in-time analysis of the base-year data, are mean income over the period had not be acconconfirmed in fte tme series. For example, the panled by any dhnge in inquaty. TiNs is ob. analysis of the base-year data nay wggest that tained by applying the terminal yea mean to the adjustment will benefit the poor, since most ate base year Loenz curve. Similaiy, let P', be the engaged in export-crop production. Observing poverty Level in the terminal year tht wowd wlat happens to poe over time in the expot- arise if mean inomes had stayed constant over producing sector should confirm (or correct) this the period, but ordy changes in inquaty ocexpectation. cured. For is, the base year mean is applied to 3 88 The P, measurs for each region and so- the terminal year Lorenz curve. The change in cdoeconomic group should be computed for each observed poverty between the two years can then year, and comparisom made. For such compai- be decomposed as follows. sons, two observations are in order. * A choice mustbemade about whether or not P 4,-P 1 O = P-Pl)+(P' 4 i-pe)+ residual (9) to be rigorously relativist in the treatment of the poverty line. Poverty line(s) may be selected for 3.91 The first term on the right hand side of (9) the base year, and kept constant in real tenns for is that componentof the chage in poverty which subsequent years. Altematively, a povy line is due to the change in the mean income, hokling defined in relation to mean incwoe might be ap- the base year Lorenz curve onstant. The second plied, in which case there is no cerinty that It term indicats that part whikh arises from the will remain constant over time. The intapreta- dcange in the Lorenz curve, holding the base tion of the results for such a case wili be quite year men constant. This analysis can throw different of course. lighton the proximtecausesofinequalltychange. - Domnance results can also be applied to Has poverty decreased mainly because of ecucomparisons of poverty over time. In this case, nomic growth, or hs thwe been a marked imthe cumulative distribution functon in the ter- provement in inequality? minal year is compared with that of the base year. If it is everywhere below the base year, Condluding observations poverty has unambiguously fallen, regadless of the poverty line and index selected. These cal This Chapter has described how the culationscanbeappliedtooverallpovertyindkces household data geneated through the SDA IS or to poverty in each of the groupings selected. can be analyzed to p t a poverty profile of 389 Having data for two or nore points in the country conened. The profile isdesigned to time cearly extnds the potta scope of the provide insights into how poverty is likely to be poverty profile in a number of ways. We shaul affected by adjustment, although as we have highlight one importat extensioni decompos- cautioned,thepfeannotprovideanyrigors ing poverty anges In growth and iquality grounds for attnibulngcausatiozl Emphashas effects. been placed on keeping fte analysis relatively 3.90 Growthandinequayefet. Thereaetwo simple. Th computations involved are staightproximate causes to any change in the poverty forward, and most of the results are presented in index over time a change in the mean income/ tabular form. Such tabulations can be easfly unexpenditure; and a change in the distribution of derstood by policy nuke Clearly, further income/expenditure around the mean. The first analysis should be applied to the data, some of is simply the effectof a general growth incomes, which is described in other chapters of t voland the second is the result of changing irequal- ume. These studies wil provide a more in-depth ity. Itisasimplemattertodecomposeanychange undestanding of the reasons why some houseover time in poverty into that which is due to a holds are poor and others not. change in mean inomtes (keeing inewquality at 3.93 The main objective of the analysis i to its base year) and that arilsl, from a change in provide policy makers with an improved empiri- 64

73 cal basis for protect poorer gwups dwing weth, sdo giaum md endm soffdiwysentl adpustment t slwtld sgl dealy in bouscalds. wherepoor sae partiularly vuln be to 11 The setim of wsdewoomle groupswed in tn poliy change, and whee the ale tn need of tabnisfferaet kom ibatpropmdabove. Nolwcam inumediate assisa But me hiprtantly, it should be rd nto this - tw eatgain the tas are should suggest ways in which poliq Inven- awewbhlweremeuinstedesacncuesweandahana. tcownlghtbeadju Inordetodampedown 12. Andabo o_dlseetheppvstyllu ected Mm adverse effects, and enhane any beneida ef- mm ouderus we obtaed using the 30h paoute cut fects of policy Inments off tse 10th perntie I& SeeAt s(19s7)andlouteandshourodsm(1). Notes 14. lbe weakar somed order domdince tst that the e,adertieamuvedaationfum forgroupab 1. The are Inddviduavel da obtdn IS, less than groups. ardng aewloymert, eduction, het nutrbion ad soe n. 1. Solongastbepoveratymes e ae ertinild But xenditur are ony obtaid fw the househdda * a mditone, nably thatih a_sqcntinu e r,w saetwhole. vie and wt_dy _momtn 2. The e_ tm of the 1tentel hbotb f epei. 16 aceary, to ethese catsheanalystmust tneasap oxylorwefilfowsdbtof Boatng1at (99. hae at hs dsposal an ewpihil exeadur syte from Appendix 2). whkhtoainmheho daoaetoptwp epdcd ges 3. Along witha h v dds o onbcd bhvor, we undermdro asum dut prductim dedshms ac be solved sepately 17. An altrative would nvove the ue of a g_ral from connpndeco See World Bk (199) for a equilbim fnwwr. though hee is some eviden that dboissuonofs ndhe sivehouddnwdds. little would be ganed hom thil In the cas of Cole dlvce, 4. Foradsionofthedifferenesnreulisobtlned Ktnbur(96 )aesthat'tas gssitoftefsedbade usig altamive nomu_ of wellfae tn th ecs of Cbwe d w via a an gen mex qulbrt m mode does not dlvoir,ssee Gewwe and vander Gsag(l) - to Inadae theepoyguf&e thatemew f fmte 5. Ourttwf*-b_b sed pwipnsilamds. mea ueasumed tath ycommodllas andewlfar 1 Weghts =ed in sumt prke indies shold IdeaUy funcio The ndan Blmitoan of ths meue Is that ft 4g- ree the ip cnblatim used In the r conceed. na utlty that is dedved bun l1m To cpue th, a It may mnot be possibe to cmpule t prce fdies or f.ull-noome approch would be awaded, whidh ags a ehoftesodslsectrs,inwbidh eproxydeaatosmit monetary vlue to lelwe wsing the opportwyge wv n w be usd (eg th anmer pe IndeW or, sne labor cmcept. priseethe maintputamoneywageidex) 6. PF a review of the ive melts and deerit of 19. Aswithbour obvaoof m fitheekoondcnhfasru_- food expenditure asa wefa m re, see Gwwe (19). ure, we thbsipole i depnd e sampe that 7. Por vefl reviews of recent de_vpnts i the do bdawn.ifutepooraedipedaau safltheammnlhies, ofsovrty, seoalbo Atknn (1987), FoPSt twblle told howt lfia_ (1984) and Sen (195 and 1967). havealcted (asdisbtfnfodm h dtlddgoups). a During the 197 poverty 4 was populry d edas Hwowev, nule the ntboenfratte sal Irs. the poomest 40, mainly ma result of the anlysis peoenbtd tuctud changes wil have noeble dects at the hawse incheneryetal (1974). hold level, and hese my reveal dfferential act of ad- 9. Although gi e ln,tesswieywmberepeatd justuntn the poor and nonpor. and obsevatis fo a namber of pdit In me wil be 2D. Pw an applcatio of this denx se Knkavalabl.e Mm_pUllos of tis for po analysy are wand(1989)fortled'lvokeandravanandhuppl(1989) reviwedinsectinvbedow. for IndonSia 10 Pystt and 1orbcke (1976) dlscs the mes of 65

74 Annex: Possible bias In poermeaurement It Is important to hgtdight a possible sed. The direction of the bias for 1P nwasues can weakness in the SDA data which may affect the orly be determined empiricaly, since a priori, reliability of esimates of poverty that are gener- the effec on P. is i de t (dpend Inated. A feature of the SDA householdlevel data ter alia, on the position of the poverty line in is the relatively short reference period that is reation to the mean). However, there are two used in obtaining esimates of some of the con- sours of consolation for the SDA analyst faced ponents of Swcome and expendture (in some cases with this problem. First, insofar as the analysis is only two weeks in the current SDA design, based on arbiy poverty lines, the absolute though this varies acoording to whether income extn of povert is not so importanl Rather, as or expenditure are being measured, and in the we have already mentioned, it is the paten of caseof thelatter,on how frequently theexpendi- poverty across the various groups which is of htre item is purchased). Although the inc.ome/ polic concer It is unely that the poverty expenditureexperienceofa household will rarely rankig of the various groups will be seriously be exacly typical of annual income/expenditure affected by this bias (uness the source of the bias for that household, by summing across house- -thatis the non-representativenesso rferenceholds unbased esimates can be btained for the period inomes/expenditures - happens to be f population as a whole (or for subgroups of the significantly wore for some income groups). population) of mean incomes and expenditures. Secondly, it is possible that whilst the However, as estimates of annual income at the m eent of income dispersion using short individual household level will not be accurately reference periods is subject to this bias, the measmeasured using the short (two-week) reference urement of expenditure dispersion will be less so period, esdmates of the dispersion in incomes (assum that both mues take he same ref- (and other statistics which depend on the disper- erence period). This is because households faced sion - for example, the Gini ratio, variance, with income variations will tend to smooth out stadard deviation, etc.) will be biased. Scott e tures over the year, based on expecta- (1989) has shown that this upward bias in the tionsof the 'permanent'incomefortheyear. The measure of annual inwome/expenditure disper- evidence Scott compiles to show that the bias is sion or variae is both inevitable and large potentiay serious are derived from measures of (posibly of an order of over 30%). The extent of income. Moreover, some of the expenditureitems this bias will obviously depend on the impor- derived in the SDA survey are obtained over a tance of those components of income/expendi- one-year refrence period, which wfll reduce the ture (for which short referee periods apply) in resulting bias. Obviously, whether or not the total income/expenditure. For example, most nmasurement of income is more prone to this food expenditures are obtined over a short ref- problem than measured expenditure depends on erence period, whereas other less frequently- how they are derived in the IS questionna If purchased items are measured over a longer income is reported over the year, then this bias reference. Similarly, some income data (for ex- would not be present at all for income eases. ample, wage income, ae obtined over a shorter For agricultural incomes, for example, the IS refence period than others) questionnaire takes a one year reference period, It goes without saying that any upward and for agricultural households, one might anbias in the measure of income variance has seri- ticipategeaerproblemswithexpeditu-based ous implications for poverty analysis, since the estimatesthaninwomenwasures(abstrcfzfm measues of poverty are also certin to be bias problems of recal error, that is). 66

75 p~~~~~~~~~~~~~~~. ~ ~ ~ ~ ~ 2 to~~~n ;3I. g i. q ,l,,,3 v *9Xl }X-...-,-.--.'l.i----ll

76 th budget defici). The pris of nontradable 44)5 Product maret adjustmts alo affect goods and srvkes fi rative to the pices of labor supply. Price increases for essentidal comtrdables, beause domestic supply and dend vkodities wil raise abor fr paipation, and determine former prices whereas tradable hours worked. Reduced unccmes among the prices are given (or sml African economies) by houeholdsf main earers will icrease partidpaworld price leves. This in tum encourages tion among other me_ms (more school-age traabes production to ekpand, and nontradables childoenmaywork,forem9e). Fiscal measures producin to conract, the speed and extent of hemselvesmay affet labor supply (for example these changes depending on supply responsive- by reducing transfers to houds and irneu ntessing their xation). * Dewiuation and Import Urraltion. The former increas te domestic prices of tradables Det laborrmaiketefts relative to nontradables which furte encorages tadae expai Not au tradables ex Ad_sment directly affects labor nmrkets pand, however, because cuts intariff and quotas though: reduce the prices of Imp bles relative to ex- * Chnn Ohe prw of prkate-tor wag-deportabl Sectorsproducingprevouslyprotected tmnation Governmentminimm--wage setting Importables (usually manufactures) therefore somedmes encourages a dual emmployment sruccontract relative to exportables whi are mostly ture; wagesc anot be adjusted downward t clear unprotected (or suffer from negative protdect) 'formal' labor makets and the unempbyed must in Africa. esort to the ifrmal sector. Minmum-wages * SctoWl prie adjusmh enx The lessnang of mayberelaxedundermarkt-lbralizonwith price controls the adjstmt of controlled pries sigficant implications for formal and omal to market-levels, and the reform of marketing livioods. instions to increase thir effiency, tgete * Cakging te pos Of publi-ctor wage-dalter the struture of producer prices. term lnsoveeountiispubllsectorwages Chages in the provision and rdraton of ae still set abov market4evels. But in others, eonomk inmtrucuw Public invest_ent budg- wages for professionals have fallen below those ets are recast in the light of new priorities. This in the private secbor. Most adjus_tm progams alters the reurns that producers derive across try to rectify these armalies tough publcproduct markets new ta sport and processn sector reform prgms. facilities hmprove prfit ty in key scor (such * Rtrenents in pubic-sector envplyment. as agriculture), whle investmet cutbacks in dis- Wher publc employmet is overextended, favored sectors reduce their prfitability. teby ntdbuting to budget deficts, retrench- 404 Theseproductmaktetadjtmtsintur men of unskie staff rolase resources to pay alter the derived demand for labor, thereby coin- mote to scamce professionals. Given the concenpletng the trnmisslonof policyeffectrough tradon of public emplyment in spedfic ocato household livelihoods. If tradables are more tlons large chages in local labor markets are labor intensiw than nontraables, then in the implied. long run employment and retuns to labor (in 4.07 Whether labor makets operate efficiently both self-employment and wage-employment) via rapid adjustmts in wages to shfts in labor will increase, provided that no labor market *i- demand and supply, is cral to adjustmet's gidities exlst The opposite result ors when impact on labor welfwe (Johnon, 1986). abortradables are less labor-intensive than non- maret _ation induced by inappropriate tradabe. The short-run outcome depends on polices can impede this efficiency, but segmhow mobile labor is; if substatia retinng is tationcanalsooccurlrespectiveof poicy. Causes involved, or if geographi mobility is difficult, indudeundontionand monopsony,monopoly then wodrers in expanding sectors will earn a employers, imperfect information, barriers to premium on their labor reuns, while those in mobiity,discdnation,and dions in capi. contract sectdors wll face a larger wage-fall or tal wakets (Mezzers, 1981). Whatever its cause, greater uneloyment than if they were mobie s tion sgifl afects the impact of But how the gains and loss are shared also output fluctuatios on emplymet and earndepends on goverunent poices in the labor ings maket itself (see below). 4A Consider, for irtncep, a fll in the de- 68

77 a'ad for nontdabls Wth inflexible, frmal- nr whih s readiy undertood by decisonsector wag the formal labormarket in the non- makers. To achieve thi analysts and polcytradabl wsector takes tde ft of the employ- mankes should agree on a stndaw ta-prana ment ntraction, sinece some workers cannt re- Han fiwework which wll compeent exdsting ta employmet by takdng a lower wage. The informtion on employmet and earnings, and bnfrmal labor arket In fte nontradable sector alert poicy-makers to problems requing public is expcted to take the brunt of the wagfall, interveons because Its labor deand fwas, and the redundant formal workersrmove into this may result 4.12 Althoug* the features of such a framein mre informal workers dropping below the workcanornlybeestabshedattheodmntry-levd, poverty bne). The formal-informal wage differ some inportant aspects can be highlighted unential theefore rises (McDonald and Solow, der the headings of 1985.). 1. ScIoeono'iCa roup s Wan Data Presentation, Gopiald maw*et efj7ects 2. CharacterIstcs of the Econonically Active Popuation, 3. Distribution of the Labor Force Between 409 Adjustment affects capital market Tradable and Nontradable Sectors, t erougo- s 4. The Stuctre of Publc Employment, hs ocum s h roe pree Ot of tho pricus 5f Urban Employmet and Migration, Ths occmgurmh decontrol of the prxes of 6. Rural employme: fam and non-farm sf capital equipment and changes in the prices of employ.ent, Impored capital equipment consequent upon 7. Lab lows between Occupatios and Secdevaluation ad import liberalization, 8. Po behnoc t as * Chang in the strct of fpul incal f- f tousersofapit Progmmoftenincludethe 9. TheRncwhnerofE arnisfromnon-farmselfreduction of tax allowances on capital p ha, Empy t, and and the dimnati of capital subsidies. Wages.. Elinion of finacl epsio In many countries loan interest rates are kept below mar Topics I to 7 aover employment, its chrket levels, thus cheapenig the cost of financing acteistcs and distribution, while topics 8 to 10 capital purchse by borrowing. The upward cover the structue of earings from different adjusthent of intrest rates to mwrket dearing employmentsources. Topics2tolOareaddressed levels is an imporxta poliy refom in turn, and sample frames (or tables) suggesting 4.10 Each of these in turn alters the relative ways to present corresponding survey data are factor-price ratios hfc both households and proposed. The identification of homogeneous enterprises, lading to changes in the factor co- s nomic groups and taget groups for sobinations used. On balance, in nost countries, cdal dimensions (induding employment and such changes wil tend to make capital more ex- earings) analysis is discussed in Chapter 3. pesive relative to labor thus favoring employment and wage growth Te strength of the latter Charactrtics of the econcaly active populaion will, however, depend on how quiddy adjustments in capital stocks can be made, and on the 4.14 Although this section of an Employment elasticity of substitution between capital and la- and Eanings Report presents the most basic inbor. formation, its interpretation can, as shown below, yield considerable insights into the employ- A descriptive analyis of employment and ment and earnings effects of structural adjusteamings under adjustment mo The SDA Integrated Survey contains a 4.15 TomieumoNoapeAcmvupoo'uAT1NBy coiderable amount of Information of use to the mwn wo amnr. Frame 4.1 (ovedeaf) shows the analysis of employment and earnings. This sec- 'economically active population'' distributed tiornshowshowthatinformwtloncanbepresented across their main emplyment, with disagggin the form of tables to inform polcy analyses. It tion by gender, age, nationality group, and is Importanto present the information in a man- whether the Individuars household is above or 69

78 Frme 4.: DsdrAbutlon of te active Frame 42labor foe prtp patimn nraio (LW populatin by main emplmnt au, ad unemploymet (U by reion, gender, ae gender and age goup, nationaty and by poor grop and household headship and non-poo households (, Female AR As abowe male AU AsabovepI; Nalianatity grops Poor Nam-POW s ~~~~~~~~~~oa LUj- Employe EMPl4Wd Sdf- uaia w*0 Ruwd Nati - L U L U L U L U p Age= Ousholdhea mb ~~~~~~~~~~~~~Otherbouseholdweinbezs AU Male Asabove AU Age groups Natosaty grops AD below the povey lie. 2 A relatively fine disaggreation byage is eomm given - variations in employment prospecis. Sinem 4.19 LAuc PAm PAI0N Am UNDeULYupatonal uit occ among the young. m Frame 4.2 presents labor force partlpathe period from 7 to 25 yea should contain the tion rates and unemplymt rates by location, argest number of age groups. 3genmer,agegroupandhouseholdheadshp Some 4.16 Frame4.1offesauseful sicpiueofthe ceus may contain only a small nmber of obserlabor foce, dcarif (for ance) the stucue vations. For example, unemployment among the of youth enmloyment, and the role of public em age-group in rural areas may be very small, ploymen Reporut Fram 4.1 at regular inter- and this should be indicated to the user, or exvals, from sucvessiveys,provwdes anamy k cluded. However, cild unemploymetmay be picwe of how employment changes under ad- very sigificant in urban areas. In some counjsment. tries orphans and abandoned children may have 4.17 For Insne in the shot run, both private migrated to towns in searh of work. Inreasing and public employment may fal heavily as a umplmentandparticipationratesamongthis result of demand contraction affecting both pr- age group wod indicate a serious chld-wlfare vate and public expenditues In tui this em- problem, incud schooli deficiences, dnce ploymentlossnmaybedls pomtelyspread; chldren in this situation are unlely to be enyoung workers often have less job-security tn rolled and attendin sdhool older workers and more women may be made 4.20 Dlsagegaton by location shows the redundant than menl Employment aftenatives situation in local labor marketsfollowing sectoal also vary sigificaly; young men may have policy changes, and how tids impact on target fewer opportunties for self-employmenthan groups. For instace, the employment effects of older men (because the latter may have geater industrial rationalization are often oncentrated aces to productive assets), so open unemploy- on parcular localities. This may have severe ment may be higher among youner ma effects on a paricular group because of the char Obviously care must be taken in maldng actsfti of the inustrys employees(eg.,if the such inferences and the useof such tableswhould factory employed large numbers of female opbepartofwidkranalyses. But hepresentao of erativs). Frame 4.2, reporting th resultsof sucthe data in this way does provide a first india- cessive surveys, wfll give an overview of locd tion of situations for hier nvestingatio In ths labor market adjustent to such policy reforms. regard, with sucoessive ourveys, twe mgitude In the first survy a ligh rate of female unemof dcnges tn cell values shud be reported. ploymnitin (say)the 'othdrub' categoy dght 70

79 be observed; the next surve would indicate frme 43: Disbuthion of Ote emiployed labor whether tat unemploymnt had persisd or not force between tradables and ndales 421 Pemale laborforce partcipation is on av- (prfnt) erage, lower than that of males because women - - have lower opportumity CostS of non-participa- pi nt tion when teir wages are low. Unemployment. rates are often higher among women because radables thir opportunity costs of job search are lower, Protected trudabes and discrimination in hiring leads to a lower job- Unpotected tradabke opening rate for thum Information is thereore Nodabies reported by gender. It is also reported by house. Nnftadable hold headship, since household heads (of either Co_me goods gender) usually have a higbher participation rate, Nontadable and a lower unemployment rate than other Capl goods household members (who are partly supported Pblic eetic by the household head). This is particularly the case for females: in Tanzania for instance femalehousehold heads have a much higher participa- Dsbution of the hborforce between trdabkd ad tion rate, and a much lower unemployment rate nontmdable setors theian females in general Finally, Frame 4.2 could also include 4.25 Since adjostmt alters the output shares marital status, since of tradables and nontradables, Frame 43 sug divorced\separated\widowedwomen generaly gests decomposition of employment status by have the lowest unemployment rates, and the the tradability of sectos The household queshighest participation rates, of all women. In los- tioair.e provides data on the industrl affiliing the financial support of their spouses, these atlon of respondent's activities, and this abows women face very high opportunity costs If they us to cassify the trade, si or n i do not participate in the workforce. activity into various tradable and nontadable 4.23 Dlsaggreptions such as these highlight categories, as shown in Frame 4.3. the situation of different groups and, over suc- 426 Labor moves betwee nontrdables and cessive surveys, show whether particular groups tradables, but labor re-aocations also occur are faing better or worse than the average. The within the tradable and nontradable categories multivariate analysis of paragraphs 4.57 trough (paragraphs 4.06 thorugh 4.08). Two types of 4.68 pursues these issues further. tradables are defined: protected tradables (inport subsdtutingindustriesunder tariff and quota 4.24 EMPLOwnRr sratus AND PARmIaPATION BY protection) and unprotected teadabes (mainly wuca1nona ATrAINm-r. Frames 4.1 and 42 can agricultual products). 4 Nontradables are disagbesupplementedbytablesshowingemploynt grated into consumer and capital goods, and status, labor force participation and unemploy- into puble services. These tabulations show fte ment by educational attainment (with further sutue of employment at any one time, and breakdownsbygender). Suchtableshaveidenti- susive surveys alow employment shfts to cal columns to those of Frames 4.1 and 42, but be tracked. the rows aie: no education, years of primary, 4.27 Two additional tables shoud also be presecondary, terdary education, and so forth. This sented. First, the distribution of employed perwill show wheher men and women with the sons resident in poor households across tradable same educational atainmnts have significantlf and nontradable sectors and disaggregated by different employment pattems and, over succes- employment and self-employment should be resive surveys, wlter these patems change (be- ported. Second, the distribution of te labor force cause, forexample, menaremoreeasilyabsorbed by occupation across tadable and nontradable into cerin types of employment than women). sectors should be given. The household questionnaire provides information on occapaonal status from predetermined occupaionad Lst (manual worker, professional etc). The resulting 71

80 table will have a formt simlar to Frame 43 with ton by genderand for the poor. Thus in a given the tradable and nontradable sects defining the eglon 10 per cent (say) of the regonal labor force rows, and occupatods defining the columns. may be govennent employees and 5 per cent in sbtae entrprs Of the region's govemmnet he srudreofpublcepym emplos, 80 per cent miht be males, and 20 per cent females, whfle 10 per cent might come 428 In the past, governmnt employees have from poor houseolds. enjoyed above-market salaries together with 431 Women and men are usualy spread disbenefits such as subsidized housing and security proporionately across the column categories bes Of tenure. But these benfits have been eroded cause of variations In local hiring praces and considerably In many countries, and the struc- partcipation rates, and differences in the educature of public employment is now changing fun- tional attainments needed by public employees. daunentaiy. Since public employment is usualy Public sector r ts are therefre likely distributed unevedy, job rerenmts can af- to have disproportionate effects on male and fefct loc einploymetseverely. Therefr, Frame maleemploynmnt. Femalesmayalsobe concen- 4A shows the distribution of public employment trated in unestablhed ocuptions, whdch further across regions. Employment in govreunent and heightens their vunebty to redundancy. sistaernterprisem shown, sincet jobrenchinmen 432 Verifng whether the poor share in pubcan fall uneverdy across the public sector. Ik employment is important information for 429 Established and non-established public- countries beginning rtenchment, since subsector employnent Is also distinguished. The stanialasslstancetoredundantpublicenployees latter has become mre prominent as established is not warranted if few of them are poor (Dernery posts have been frozen, but 'temporay hiring of and Addison, 1987). However, the poor may be staff habeepennitted. Unestabishdjobscary concentrated in the unestablished categozy, in few fringe benefits, and the SDA survey yields which caw assistance should be targeted to tiis sfficent infomiation on finge benefits to iden- group. Successivesurveyswillalsoshowwhehr tify unestablshed jobs. Unestablished jobs usu- the poor have been disproportionately affected ally bear the brunt of retrenchrents since the by retenchment. To date only circnstantial unsldlledareconcentratedinthesejobs,and their evidence has been available on this issue (see costs of dismissal are low. Addison, 1987, on The Gambia, for example). 430 The table also disaggegates the iforna- Ur&m employmen and migration Frame 44: Shaes of govement and state are changes can be expected in urban enterprise employment in regional labor livelihoods under adjustmnnt since nontradable forces shaes of established and unestablised servics, and proected tradables, are urban fopublicsector workers hn the national labor cused, and these take the brunt of demand defiafo* to Fonral job losses add to the informal secor, (p_uta wich may alrady be absorbing lare amounts - -h of labor if fotrmal ecoonomic activity is stagnant Egioc N prior to adjustment Frame 4.5 therefore decomposes urban C S G s E U employmt (both wage- and sef-employmt) -A- into its formal and informal pars, using critena All such as whether survey respondents have an cbqu=o pbuc eplpoymer empbyment ontact, whether mnimum wage Female legisation applies. The table indicates how un- Male portnt fonral and informal emplyment is to Pewww from poo hwdibolds diffent age groups and to males and females. Shaes for the poor and non-poor can also be ported Fnally, the wage-and self-employment components of the formal and informal employ- N add umore muybddgvin.ge ment categories can be reported (although there 8FiUt,CUfGd5 may be insufficient observations to report ths 72

81 Frame L5 Distrbution of rban emplyed Fme4 Rlgoam anddnaracerw lbor fore betwee fom and Inomal stis of migrants seetots by ae, gende and household had- 4b.d C o _bo M AIL MU MY F I F I AV Over24 Oiw bowebold mtembers L AU howlidd heads SL}~~~~~favraphoowdl M w s inie M 4lab tes lrffb As above AU Frame 4.7: Shaie of th employed labor force AgeGroups that rerts non-f s-pl etby AU region, gender a and nionalty Nbo* iatesfipal.ydo &epo Other CAPft wbo NRol Nd infornationbyagegroup). Changesinth StrUC- e tum of urban empoyumet can thn be monitoed AF8_p as policy reforms are Implemented. Household head 435 Becau adjustmt causes substatial oth change in the ewonomy, a reversal of past ftier~~~~ Inter- Members AD y X gion migation of labor may occur. Spef- i, cally, labor may retrn to agriculture, if (as IS M. wliy) adjustme shifts mt urban-r:ral Iwnome Asabove d" 1if in favor of rural eooyn AU 4.36 Frame 4capbues prtt dimebsions A of migrafion under adjustment. For each region, Az, the dse of \recent migrants (moved durtig the Naups astyear)intheregiosworkfomceisgivn. The AU SDA survey provides reasons for migrion, and the share of rect mgants reporting redundancy or busin failure is given. To estalish ae re likely and able to mgrate. If sample how wel they have adjusted to the loal labor size permits, then dernmpositon by gender and maket, the percentage who are carrintly unen- poverty criteria should be used. ploy is shown. inay, the income of housholds in whdch recent mgrants now reside is The stuctwe of nonm sdf-epoymd reported as a pe Fentag of the region's house hold inome. Tlds also indicates how hy have- 438 Frme 4.7 shows, by region, the shae of adjt to the lox labor market If su ve thelaborforceinnorinl ent or survys are available, then their pogms can be exaple, X per cent of al empyed males (and bucked; over time theirunemployment rabtemay Y percent in the capitl) may report on-am fl and ther househols incoe may rise rela- sef nmploymet. Since tis employmet usutiveto the average, ally beames more mpoat as people grow 437 A smll sample of migrants wil limt de- oldr,decomsibgegrup shown(see composition by persal h cs Frame Vjverbea, 1988b on Abidjn for instance). Gen- 4.6 thereo gives two agrus onr, the cut- der difference an usaly evident, as are differ. off ageisthemid-twentiesinceyoungerpewple ences arss urban entrs pinay In mst,,, ~~~~~~~~~~~~

82 Frame 4& Distribution of household non- Frme &9: Cunrent occupation tme-use in farm enteprises a acvitwes by reglon elati to previous occpation meuse and for poor and no-poor households (P"=ka) - ~~~~~~~~~~cw -srqwu -***l. E H S O1,02... e AH NP P AH NP P ' - Mam SvI H S Food comerow 01 NMI-food oounewce 02 Aftindicatsallhonseb.ld. NPidate. nan-paacousetobo Pi dahctes pow hausbcid, a h o~~~~~~~~~~~~~~~~daasoskefg graicdfatim echwm I imeducaton..~~~~1... Slnlaeseriigorwok outries the are significant differees between O1V4t. All teso = nationality groups in their reliance on income from nonfarm enterprises. 439 rame 4.8 seeks t capture the importnce mead down its olumns. Thus for each present of different non-farm enterprie acss regions; main oocupation the percentages of those who for instance services are gen y more impor- were previously in full-ime education, unemtant in urban than nl areas. The SDA house- ployed, in the same occupation, or in other occuhold questionnaire gathers infomation on up to pations would be shown. three household enterprises, and these can then 4.42 The table indicates the flow of perons be aggeated to form ihe table (see Vijverberg, between occupations as adsment takes efct 1988b). The tae shows the reative roles of For example, we can see how govement workdfent non-fium entrpnses in the employmt ern are re-employing themselves by cheddng the of poor households. They may predominate in pementages of urban businessmen and famers services and foodcomnenr, where capital re- who were formerly in government. The table quiwrem s ae small, rather than in manufactur- could also be disaggregated by geder, to caping which needs more capital We can track how ture difference in the occupational mobflity of this cnges as proects are implemented to im- men and women (due to bas to entry affectprove their busi _sse ing women more than men, for instance). Fi Adjustment will cause changes in the naily,ffomationhouldbereportedforthose shares of thse activintes over time. For istac in poor and non-poor households to highght seri and non-food commerce usualy become the difficulties faced by the poor in achieving less importnt since these are predominantly occupationalmobility. bnnadable activities, while inf maufac A table similar to Frame 49 can be preturung,beingpedominantya tradable, wilmost snted wit occpations aggregated into groups likely increase. Areductioninrent-seelngupon of formal wage-enployment (with sub-catetote remon l of market distortions will also re- ries of public and private), informal wage-emduce fe role of nn-food ommerc in the short- ployment, non-farm household entprise, and tem Again the table should capture these ef- farm enterprise. Each of these can in turn be fects. dis Igated for various categoies of tradables and nontradables: for instnce using a three-fold labor flows bdween occutrn and seams castion of tability, formal wage employment in the private sector can be split into unpro- 4AI Frame 4.8 clarifies the occupational tected tradables, protected tradables, and nonchanges that occur under adjustment, using in- tradables. The quesonnaire allows the matchformation on empkoyment histoies. The curent ing of occupation with industriwal status for both main occupation or other time use (housekeep- past and previous main employment. trg, education) is cross-tabuated with the previ Such a table would, for example, show the ous main occupatin time use. The table is then pectag ose wage-earners currntly in 74 t

83 unprocted tadables who had previously been Frme 410: The structue of houseold wage-earners in protected tradables or non- Income In ruril and urban ams by poor and tradables This provides importint infornation no-poor households, and gender of house- (especaly when successive surveys are avail- hold head able) on the short and long-term mobility of labor (p_ttage) between tradable and nontradable sectors which, -- it was aigued i paragraphs 4.03 through 4.10 isaa major determinat of the labor welfare conse- Pworwde_ =m-,parmwhno quences of structural adjustment. The smci t of earngs Fe6e- MAle- Fenk- A&lIheaded 4A5 Sources of household incom Waes-emplonamu tional earnings across tradable and nontradable b*t secto, and tabulations and indices of real earnings are the subjects of this section. Frame 4.10 Sdf-ezploYmeet decomposes household income by source for poor A8dtw N'on-fium and non-poor households. 5 It night be ob- E,ps served, for instance, that uneaned income accounts for more than wage-employment among Transers poor rural female-headed households (reflecting Private their dependence on remittances) or that non- farm enterris make only a small contribution Nd1 M.acdmnadd to 100 pcel to the incomes of some poverty groups. Successive surveys, together with other information sources, can map hanges Insources of income as Frame 4.1L The occupational earnings in adjustment proceeds. For example, self-employ- tradable and nontradable activities ment income could fall among urban poverty (i lbn cwmwy oer aone monhk pw groups if policy reforms depress the informal sectr. Suitable poject interventions can thenbe implemented to deal with this problem (and tabulations from scessive surveys will highlight how income soures change as the projects are Tradabes Manjj AM= A& MW.iV implemented). Prtected taaes 4.46 As stated earlier, every employed person Unprotcted utadables can be dassifed according to their occupational Ntda,b1es and industrial status, and each industry can be Non_adables further lassified according to its degee of trada- Consmrgoods bility. Foreachuocpationindustydcluster,mean Noes and median values for eamings can be calculated and reported in a table such as Frame Given PCitAl 9gods the likely skewness of the data, it is important to Avera by o=paon report the median value in addition to the mean. If repeated surves were available, it would be posle to see how the structure of occupational and workers are abletomove outof nontradables, earings hanges as measures such as devalu- the differential in favor of manual workers in ation Impove earings in tradables relative to tradables will fail. nontadables, as trade liberalization alters the 4A7 Price indices using nform in from the structure of eaniungs within tradables, and as SDA community surveys should be used to calearnngs in nontradables are affected by mas- culate the real eamings provided by different ures such as civil service reforms. For exmnple, employment sources. For example, the mean adjustment might raise the earnngs of manual and median values of real eaings derived from workers in tdable secors, but depress their wag-employment (both formal and informal), eandngs in nontles But as time passes, selfrenployment(bothagrlcultureand nonfam),. 75

84 . 3on Frame 4T Inug of household non-faww shsuce of 5wninssfrom non-fr sdlfenterprises Lo poor anhd non-poor boushods em#plmen 4A9 The prfitabiity of hosehol entrpri E is murd by net revenues. Under adjstment G.wssvtm ijdivwug these are affected by product- and Input-pre movements (due to deoand effects, exhangrate Pr Nms-r Pm Norpw adustents, canges in tade intervtins, and - ithe relxtion of price contos). Net revnues AdE MD hee MI MA MD ME MD are calculated fkom inormation on the values of outputs and Inputs. Frame 4.12 reports these for various enterprise categories, and for poor and Foodcam_eo non-poor households. Successive surveys will NOPvDcd show how profit Inomes among the poor fre as adjstmntproceeds and as projectbased interventions to assist target groups take effect Medk diw. ydian values should be reported since evidence to date hows a skwed distbution of earnw anong most categories of enteprise. Gross revemies are reported as well, sice changes in these Frame 4e The stucr of wagedifferen- are indicative of demand conditions facing tils by educaton level adage olds. Frame 412 proposes a way to de- Cmwiu of mm nuig_s.eic) cmpose activity eani by region, ths captwring changes in enterprise profits dependin their locatio For exmple, the demand for, C and profitability of, urban services may fall f FmW 00w ute,q P,uiy adjustm substantialy redues urban fomal 3AmfkmhwM Ax w&= Rwu Pw's incomes. But due to gwwing agicultual incones, the demand for rural services and their AVV=p pwfitamty may rise. Ehqmeduti Apranps The structure of uages AV gwrp 4.50 An impartnt task in the anlysis of earn- TftiaY e'w'* inegs is tq esbid the de nts of observed differences In wages between individuals. What are the respective roles of supply-sde variables (forexample,educationalainme post-cdho, exeie) and demand-side varables (hring practies, diimination) in detrning obsered wage-difeentials? The gneration of an answer to this question requires multivarlate and rnsfes can be reported for each gion aalys (see paragraph 4.69 through 4.73). But with breakdowns for poor and non-poor house- suitable cross-tabulation of the data can provide holds, Over sucoessive surveys these will reveal some guida. changes in the stucture of eardings as policy 451 ThewagediffertsofparticularpoUg reforms take effect. interest are those between genders, between 4.48 lasty, given the effect of adjusmnt In regio and between the private and public secchang the formal-nrmal wagdifferentials toa All of these are lkel to chane under dicssed in paragraphs , it is adjstment Frame 4.13 rports on these differotato to povide real sfor both entials pessed as ratios bewhveen, for Instance, fomal and Informal worker Wage disperson mean female and male wag-rates. In sorneoccucannarroworwidendurngadjustmentdepend- paionsworkseivefrnigbeneltintheform lu on tle nawe of the polcy padkage of health, ung_ and trasport sudies. Earn 76

85 ings-differentias. inluding both wage and non- A multivarate analysi of employment and wage payments should twrefore be reported. eamings under adjustment 452 The rows of Frame 4.13 are distinguished by educational level, wit further sub-divisions e dertminan*s of labor supply in terms of age to proxy for postcol experience. Ifwages are maidy detenmined by human 457 PAIWxAOTI AND HOJ C WOI With capital and work experience then the ratios will multivariate analysis it is possible to go much be close to one. How far actual values diverge fuherthan thecros-tabulationsproposed above from unity provides an indication of demand- in understanding the impact of structural adjustside influences. For instance, goverunent wage- ment on labor supply. ITh determinants of both setting may generate a significant public-private the decision to participate in wage-employment sector wage-differential. Itmaybe found that the (or not), and, for participants, the numbers of publhc sectorpays above the market rate for young hours supplied must be established. A dichot persons with primary education, but substan- mous dependent variable LP, whose value isl if tiauy less for those with tertiy education. the psn participates, and lf they do not and a 453 Regional wage-differentials may exist be- dependent variable HW (hours of work per week cause of geographial imniobilities, so that inter- or per month) must be defined. The equations to regional labor flows do not equalize regional be esimated are wage-rates. 6 The differential between the 'capital' and 'other urban categories may be closer to U' = f(w, Y, T, P Z,X) ) unityforthosewithteraryeducation,butgreater than one for those with no education. Analyzed HW - f(w, Y, T, P, Z,X) (2) further in the light of what is known about the operation of urban labor markets, it may be found where w = the market wage, Y = profit incomes, that the uneducated face a particular banrier to T = transfer or unearned irnome, P = an index of entry into the capital city's labor markel consumer prices, Z - a vector of the characteris Gender discrmination in hiring practices tics of the household and individual, and X -the wiu create a significant male-female wage differ- chaistics of the conmunity. Similar exential, by definition. The male-female differm- planatoryvariablesareusedinthefirstestimation tial may be dose to unity for persons with secon- of each equation (although the final estmations dary education, but much below unity for those wiu differ when some variables are insignifiat with rno schooling indicating - perhaps - dis- for one equation but not the other). Separate crimination against uneducated women that equations can also be estimated for males and needs further investigation. Comparisons of the females; individuals from poor and non-poor differentials over successive years can reveal that households; and rural and urban workers In a particular facet of adjstment is s gthening this case the dummies for these chanatei or weakening di nation, are omitted. 455 Depending on the sample size at hand, 4.58 Frame 4.14 (overleaf) presents in more defuther disaggrations of the columns of Frame tail the types of varables that can be included. It 4.13maybepossible. Forinstancedisaggregating was argued earlier that age and marital status the public-sector into goverrunent and state en- signifinfly affect labor supply, so these are intelprises may reveal that a substatial public- cluded. Education is likewise a standard variprivate wage-differential is due to wage-setting able in such equations. Time allocated to both in only part of the public sector. Or disaggre- housekeeping and the care of young-children gatingthefenale-maledifferentiabetweenpub- limits both participation and hours worked lic and private sectors may show that the differ- (Grootaert, 1986:164). Since the burden of these entialis concentrated in only one of these sectom ties usuaby lies with women, we can 4.56 Finally tables summarzing the character- expect these variables to be particularly imporicsof householdsderivinga wage-income from tant in explaining their labor supply. Variables sectors covered by mnimum wage legislation from the SDA comunity survey are also inshould be presnted. From this it can be aseer- cluded, since poor peopws access to health and tained whether minimun wages carry benefits education fadlities may limit participation by for the lowest-income households or for particu- raising the time spent m, and the need to spend lar targetgoups. time taldng children to remote health facilities 77

86 Frame &14c Wageep Indend bor to houselold entrprises, and to ansume varables edecs to goods and servis. These thre sets of decisin patilpate and (1) the nmber of hours are best viewed as the outcome of the same suppied by partpants bousehold manimizdton proces Ths b why pmrfit income from household enterpses (both Kay nvdiu of uin bla d farm and non-fm) and the prices of consuner Wars Wapnteperpodod goods are included in both (1) and (2) Adjustment will simultaneously affect Profit Inane PFit oftam esprnes wagrates, profit hwoms ard mmdity prices, Profltofn<on4amenterprlsu as discussed in paragraphs 4.03 through 4.10, and the household wil re-allocate the labor of its lh_itsf Hnumembers accordiny. Change in wage rates comeorpd<nes <;3Xlamm psdmxexk caused by policy refomns wfll not ordy have the usual income and substitution effects on labor- Age supply, but will also affect the latter through mt sth UbdWT changn househld profits where the houseold use hire labor. The latler effect ndpgh be larg Dv_dowed in the ura infonnal setor, but small in ural Spoueof houshdd head areas where households use leas hired labor (this is one reason for estimating separate urban and EduRaUOIYa1e=r ofpyerdiool1ng rural versions of equations 1 and 2). Years of =ondary smdwaug 4.61 Profit income is affected by changes in fesr of vocxmv d Ag Ye=*gh%her e&atiuo produaw prices, input prices and economic in- Yeamn of schoog ampled rastructureunderadjustment. We would expect Semdaryshooliebngiped profit income to be negively relabed to both LP E d C 04 of year ddktm S9ye * and HW, because a rise in profits will cause homswholds to realocate their labor supplies from di- s ^wage-employment to self-employment. The esti- Ho eworkvaiabls lme set olectingfieood mated equations can be used to simulate the eflimespentailectlngwater fects of (say) producer pice inreases on houselrme devoed to ote housewrk hold profits, and thence on labor supplies. NWWnitty grou 4.62 The estimated coefficient on the P variable shows the impact of changes in consumer prices otn Uban Oherurban (following devaluations, sectoral price adjustments etc) on labor Supplies. 7 The size of this -AVDI coeficient can b expectd to vawy depending on dume ariable O cwe fadlity wm the the share of onumpdon met by ownt-poduc- amtmidty tion. Take the example of a food price dse, for Heath-re filty wit the itance. Non-poor urban households might be Omly able to participate more, and raise their work bulm of awn developnient 1 of det area hdurs, td meet th Ngher Dst of their consmp tlon basle, and presrre thir nutntionl But fth coefficient on P might be very low for poor households, indicating possible cnstraints on etc. Lilewise an index of area development us- their labor supplies, wich endanger their nutriing cnmiunty survey data i included. The tinl For rural households, food may cqy a estimuted equatios an be used to assess poliy- very small weight in their consumer price index, induced dhanges in these variables For insace, if mostare food produes But eirlabor supplies the impct on urban female labor supply of pro- wil be affected (but in an opposite way to that of viding nr education for females, ild-care- urban households) by the profit effect of g*her ciiiiies, and Inmroving community soa sev- food prices. Again, the separate estimation of () icecanbepredicted and (2) by sodoeconomic grups is shown to be 4.59 Dedsions concenng the supply of wae-. important laborare i endentwith decisons to apply 4.63 Unearned income C) Is indluded to cap- 78 i

87 htre the effects of chnges in both private and ket-wage. The corrctionof tisbas Involves the public transfers to useholds, both of which are estimation of the Mills atio' from the probit affected by adpjstment. For example a fall in esttion of the partpatin equatiom (1).' wban to ua remittances (due to adustments The invere of the Mis ratio is then inserted into impact on urban nontradables) might lead rua equation (2) and OLS Is applied. househols to compensate by increasing their 4.67 A Ibid ecnometric problem is sinultalabor supply. A fall in public tnsfers (e.g, food nerty aing from the fact that the wage-rate in assistance) as part of budget cuts could have a (1) and (2) Isnota truly exogenous varile,a since simiar effect, or through dminidshing nutrition, it is demined by labor demand and supply. If acually reduce the hours supplied by poorer uncorrected, this leads to biased estimates. The households. The latter is a parclarly inmpor- solution is to use the predited values from the tant effect to capture for policy purposes. estimated wage-equation (3) below, and to insedt these in pb of w In equations (1) and (2) EsmAo mumn Both (1) and (2) can be 4.68 Overall, SDA aralysts ae advised to exestimted by Ordinary Least Squares. However, periment with a variety of funcional fonm and a numberof importan econmetric problems can esiation procedures to debrmine the most costarise. s The first of these concerns the choice of effectivemethodsfor obtanngparametervalues whether to participate or not. The estimated ver- of policy ints_ sion of equation (1) can be interpreted as the proability that an individual will partcipate in The dedermnts of wage-i5gs the labor market, given information about the pesos dcarbtics. However, although the 4.69 T1i EmRoG FUNCi. The key tool to underlying probabiity model might be corect, a analyze wage-differendals is the earnings funcgiven sample of observatons on the independent tion, which in general can be expressed as: varibles may contain some observations which are assocated with LP values outside the [0,11 W f(s,e,z) (3) interval This mneans that if we used the estimated version of equation (1) to predict the val- where W - the wage rate over a hosen period, ues of UP, some of those predicted values would S = educational attainmzen, fall outside the [0,11 range. Thus, esdmating (1) E - work experiene, byois effectively constains values to lie within Z 3 charac of the hndvidual and the [0,11 range, which nay lead to biased esti- household. mates of the coeffcients A standard solution to this problem is to apply the Probit technique. This essentialy trnforms the original model in such a way that the predicons wil lie in the [0,11 interva for all ame 4s: dent vadabls hr an values of the explanatory variables.9 The tech- eam4.p hunden niqueisiomputaionallymoreinvolvedthanois MWUIU =fii t Wflc raou ProUbit uses non-lnemaximumwlikelihood eswimation). Pesearchers should dhoose their tech- Key lhs n9y W Md.I nique after comparing the OLS and Probit estimates and establishing the size of any bases in- 3duc Yem of pimydoo I&fg volved in OIS, and thus the gain in using Probit. Yef 9cd3soo 4.66 he second econometric problem, namely Whe UDiVenimy SUMPOf the problem of selectivity bias arising in the esti- (dum) mation of the hours of work equation (2) is wellknown. Essentially, the sample on whicheua- woe* Yealabrmv O tlon (2) is estimated is self-selected, given the prior dhoice whe to work or not, and is there- MSeWrIO* fore non-random. This leads to biased parameter (4ummyvutaid, Locatin (bn.url et estimates when OLS is used. The problem par- AprlcUlarnot iulaalyaffesthelaborsuppliesofwomer,since PubllCeuqympt their reseation wage often lies above their mar- Non-m entepfse or not 79

88 4.70 The indusion of education and work ex- represt both returns td capital and labor. The periene vaables is standard to earnings fun- resulting OLS estmates suffer from the selectivtionsderivedwwitnahumancapital framework.' ity bias noted under labor supply abovem sine Frane 4.15 shows the variables that a tpial individuals can choose betweh wage-employearrings functio might include under each of ment and self-employment. Hece, the sample of the above headings (Chiswick, 1976, Grootaert, wage-eaners is again non-random or 'self 1986: 203). The dummy variables for location, lected' (Cbdswkk, 1983). Various solutions have and occupational status (public or private, for- been offered, incuding using a dummy variable mal or informal etc) are indluded to capture the for activity status (Blaug, 1974), or the share of effects of wage differences due to market seg- self-employment income in total income mentation (whkh is, however, more fully ana- (Chiswck,1976),buttheseareunsaisfamtoiysine lyzed below). the varables ame not truly exogenous. The best 4.71 In addition, separate earni functions prcedure is to indude in the estimated version should be estimated for urban and rual workes of (3) the inverse of the Mis ratio calculated In the case of rural workers, the indusion of an from the Probit estimate of the participation index of producer prices captures the effects of equation (1) above, and thenesimate (3) by OLS. changes in labor demand on wage-earnings (Alessieetal,1989). Thusinanestimateforrual EpIWlaining UWe-erningse he publivwte wagewage-earnings an index of producer prices fac- d&meenfil ing farmers would be included, while for infornal and formal private sector workers, an index 4.75 If labor markets are perfecly competitive, of manufacturing prices would be used. workers with the same chrac will earn 4.72 The amount and quality of the productive the same wage regardless of their sector of emassets possessed (or rented) by the household phyment. However,ifsomeewiployerssetwages nay be positively related td the wage-ea of in excess of market-clearing levels (and restrict its members (Sahn and Alderman4 1988). Those access to jobs in their sector by rationing), then who derive non-wage itcomes from productive different wages may be observed for workers assets may be better nourshed, more productive, with the same haracterstics The earnings fiucand therefore more valuable to employers, than tion (3) described above repreents a first step those without (Dasgupta and Ray, 1986). Meas. toward the analysis of wage-iffrentials by i- ures of household assets can be entered into wage- corporating variables for enterprise-ownership equations using infomation from the SDA ques- (public sector or not). Significant coefficents intionnaire. dicate the presence of labor market segmentation 4.73 Measures of area development, derived between the public and private sector. Simlarly, from the SDA community survey, may be in- thegendervariable sindudedtocaptureposslble cluded in separate wage equations for rural and disciminationatwomenamongemployer urban workers since these positively affect wages 4.76 However, one can thus go furhr by esti- (see Bardhan, 1984: 49, for example). Measures mating separate eanings functions for wageof community access to social infrastructure eamergroupsofpolcy interesl Attheendofthe should smilarly be used, sice poor health-cae, proceaure we can thus report the percentage for exanple, wili reduce productivity and there- shares of obseved wag-ifferetials between fore wage-earnngs Theestimatedequationsmay sectrs due to differences in worker characteristhen be used to predict the eins effects of tics and difec in wage-structums In the projects at the household and community levels following example, eamning fhuncions would be designed to improve economic and social infra- estimated for govemment, state entrprise and structur They also reveal the loss of wage- private sector workers, where the subscripts deearings (espially among poor groups) which note the type of employer (g = govemment s 8 occurs if such infrastructure is cut back during state enterprise, and p = private sector). The adjustment. explanatoiyvariableswouldbethesameasthose used in (3), with the exception of the dummies Correcting fo sdetivity bias for entpise ownersip. To facilitate the exposition let X stand for au the explanatory variables 4.74 EamWstudiesgenrallyexludetheself- used. Three eanings functiokn would therefore employed from the sample, since their earnings be estimated of the geeral forre 80

89 W -fpx) (4) DmW,-Wp m g-f, )+(fsx-w,) (9) W.= f.jqc (5) E + U Wp =f,os,) (6) The first tem (E) of (9) s the part of the differential expla by private and govemnt work- Earnngsequation (3) implctlyassumed that the er having diffirt ch while the secthree sects have the same wage-stuctue, and ond term (U) i the p d differn bethat wages for workers with identical character- tween governmnt and private sectos in their isticswilldifferacrossthesectorsonlybya murk- wage-funco If private sector workers had up, captured by the dummy variable for enter- the sm chatrics as governen workers priseownership. For (3) not to under-esdmate then B would be zero, and any wage differtial or over-estimate the effect on earnings of enter- remain would be entirely explained by differprieownership, the three earnings-functions ences in wage-seting practices between governabove wouldhavetodiffero ly in helr intcept ment and the private secor as caped by U. terms,and notin their xefficients (le. they would Alteratively, if there was no sectora difference all have the same slopes). This is obviously a in the latter praces then U would be zero and restrictive assumption, and the SDA data set al- al the observed wagedifferential would be exlows the analysts to estimate the three eain plained by differ in worker characteristics functions separately, thus improving the accu- such as schoohig ray of estimates of labor-market se tion. The usual Chow tests cn be employed to test Wge fiai Gender,finn*nnal,and whether the differences in the esimated coeffi- trad/nontrade dents betwem the thee equations are statisticallysignificant The procedure above can be applied to 4.77 Having estimated (4)4(6) above we can deconpose the observed nale-female wage difconstuc lepresentativel workersforeachenter- ferenmal into that caused by employer discridprise-ownership ctegoy. This I done by calcu- nation against females, and that caused, for exlating the mean value of each of the indeperdent ample, by females having less education. If disvariabes for all the workers in each enterpse criidnation is the main cause, suitable employcatey. Each of these mean values is multi- mentlegilationisneeded. If fitiduetd different pled by the espective estimated coeffient of charactistics, then fenale education and other (4)46), and the sum of the products gives the crtics need improvement. Policy intpredited wage for the representative worker in that ownership category, ie ventos in these areas may be of great impor- tance during adjusutent, if women are to be enabled to benefit from new labor nmrket opportuwe W-W.iWp (7) rdties Private sector wage-eaners can be cassi- The next step is to calculate what the predicted fled into fomal and informal, and the wage-difwage wwould be for a presentative worker from fierental can be decomposed acrdingly. Some (say) the private sector if he/she were paid ac- of the 'unexplained' diffetenial may be due to cording to the government wage stuctre (fg), formal sector employers setting above-market it. wages if this results in profit-improvements through higher worker-productivity (te 'effi- Wp5 f.09 (8) cdenry wage ament). Altetnatively, workers and employers may havean 'imnplidtcontract' to In other words, Wp Is the predicted wage for keepwagesutablewhendemmndfluctuates. Both someone having the average of effiiency wages and implicit contracts may ocprivate-sectr workers, b subject to a govern- cur in the foral sector, because formal enteraent wage strucw Tus the differe (D) in pdses have lower turmoer rates, and use more wages paid (for example) to representative sied workers than info l entprises (dnce workersinthegovmmtentandprivatesectms is formd capital-intensities are higher). These decomposed as praices are unlkely to occur in the informal labor market where labor is often hired by the 81

90 day, and where waes are expected to adjust tected ftadables. Once employes ce Intnarapidly to dhages in maket conditions without onl compefion fteir monopoly profits would the lags which ocur under longer labor onbac. fall, thereby reducdng their abity to pay workers 4.80 Identifyingdifferencesinnformal/nrmal more than in unprotected sectors. wage-setting procedures within the pnvate see Much of the informal/fornal wage differtor is important, since rtey may intensify the ential may be due to wage-determinaton in only dual labor market structure (unemployed work- one part of the private sector (e.g, nontradable ers cannot bid down formal wages and must monopolies). The differential between each part therfore resort to informal employment), and of the fomal private sector and the informal secexacerbate the Costs of adjustment to informal tor should be estimated, and again the differenworkers. Decomposing the informal-formal tial decomposed into worker characteristics and wagedifferential on the basis of data from suc- wage-setg practices. cessive surveys will reveal whether adjustment has caused formal employers to become more The detemdnat of the eamings of non-fam flexdble in their wage-setting pracices (in which enprises case the fonral-infomul differental nay fail), or wther rigidities have persisted (in which case, Profit functiom as paragraphs 4.03 through 4.10 stated, the differental will widen since infomal wages are ex Profit functions are used to analyze the pected to fall disproportionately), detemts of the profits of household enter Oneofthereasonswhyformalenterprises prises. Let P stand for the net revenues of the can somedmes pay above-market wages is that household enterprise (which is gross revenue they are monopolies in domestic markets, therby minus au variable costs). Thau pmitting managers to share excess profits with workers (whose bargaining position is strength- P f(l, K, X) (10) ened by the lack of competitive pressures). For nontradables, limited market-size may cause where L is household labor, K is capital stodck, 'natural' monopolies, while the lack of foreign and X represents selected household characteriscompetition to bussses in'protected tadables tics (Vijverberg 1988a: 5 and Chswkk, 1983). u encourages sinilar monopoly posiions in small The coefficient for the capital variable is e redomestic markets for importables. The latter tum on capital, while the cefficient on L is the cause has become more pronounced in recent returnonlabor. Fame shows in more detail years as governments have extended import re- the types of explanatory variables which can be strictions as a means of cutting trade deficits. used. 13 They are, however, likely to become less impor Household labor is a variable constucted tant durng adjustment, as trade regimes are hb- by aggregating across working hours by houseeralized. hold nexmb, drawingonquestionnaire section 4.82 To captr thee effects on wage-detenni- 4 part B which gives hours per week worked. nation, analysts may categorize private formal Since the dependent variable, net revenues, Is wag-earners according to employment in pri- given per month, household hours worked per vate tradables, private protected tradables and month should be used. The survey does not private nontadables. It is suggested that earn- collect the hours worked by non-household ings functions be estimated for each category, nmebers, so this is proxied by the number of and thatthepditivetestoutlined inparagraphs hired laborers and apprentices working in the 4.75 through 4.77 above be appied (for Instance, enterprise Altnnatively the monthly wage-bl predicting the wage of the representative worker for hired labor can be entered. However, since from the unprotected tradables secor if they were apprens are paid only intermittently, the wageemployed in the protected trdables sector or in bill can be an imperfect proxy for hired labor nontradables). The proportion of the wage-df- input. Both owned capital and rented capital are ferentlal which is not explained by worker char- entered. These are the aggregate values of buildacteistc, gives an estimate of the effect of dif- ings, equipment, machley etc used by each enferent product market structu on wage-set- terprise (te survey collects this infomation by tig This indicates the effects that importe enterprse). alization would have on wage-anings in pro The dunmy variables for governmnt 82

91 registration and contracts for hired workers are Frame 416: The detevulnaut of ihe prfits of used to captue the effects of enterprise fomal- household non-fern entpises ity/informality. The survey also identifies who the enterprise sells to. A dummy variable for Kym " DEAof LK*fMdd foreign and doestic buyers Isb nluded to proxy w Totl w for the efect of whether the enteise is tradable mu workers in hentete or not Buyers can be grouped into either formal for a given peiod) (public enteprise, large private entepise etc) or informal (other households, local trader etc), and n4 hlbm Num of hbed bs ad dununies entered for these characteristics (this e crepras w captures the benefits of supplying the formal sector market which may be less competitive than Own capita Total vaie of owned cap the informal market). The location dumnmies stock wed by entepe (in local capture th effects of diffrences in local markets nue9cy) on enterprie profits. For instane household enterprises may face more competiion in the usetba enotepdse(in lor capital city than in rural areas. cncy) 4.87 Household characteristics include the gender of the household head, the average edu- Goverment registration II the enterlse is registered cation levels of household members engaged in wfth a govement agency, Oif the enterprise, exeriene and nationality (the latter eflects thedomiane of foreigners in some Empymentcontracts 1f1 hied labor has frma ifomal activities in West Africa). contract,o ff not 4.88 Profit functions should be estimated for the main types of husehold enterprise: food urcfhaorsdomesr' c IMa,Oifdomestic commerce, non-food commerce, services and manufacturing. Estimated coefficients can then Foral Ihormal I If formal Oifformal be compared to highight differences in the struc. puer ture of each activity. Since food-commerce and non-food commerce together with services are Locaton ifuba,ol not primarily nontradables, while manufacturing in- Uan lot I ff Capitl cty, O ifnot cludes tradables, odmparison of earings funcons esimated from sever alsurveys can indicate Businmage I iff Ild businss 01ff 'new the effects of adjustment on the non-fann sector. (aita' Ailbet of ywus 4.89 Separate profit functo (for each activ- opelfatd ity)should beestimated forenterprisesin (i)male derof head lffaa 011 not and female headed households, (it) rural, capital and other urban households, and (iii) poor and Education variables Varables for number of years of non-poor households. Regarding (i), esimating pmy, secondary, trtwyand profit functions for male and female headed vdoduton for households, gives a clearer indication of the dif- for ote house_od wurk ferent constraints facing each across the activity range. Thus theprocedureadoptedinparagraphs xerience In this Vaiable for household had 4.75 through 4.77 above for estimating the effects o ul"'m and averages (years) for othr of discrimination on female wage-gearrigs can houwetddwmnbers be adapted to self-employed earnigs. This in Age Variables for agevofhoehold volves estimating a profit function for enterprises head, and averag age of othes in male-headed households, and then inserting mean values for the indepedent variables of NaStIIty I ifnatlorial,if not female-headed households to predict the profits of the latter if they hfced the sa m e teprise earnings of male- and female- headed as male-headed households. This may reveal houseolds in services, once charceist are significant dffenes across the activitypes: controued for. But there may be substantial diffor example, there may be little difference in en- ferences in manufacturing if (say) male-headed 83

92 households have favorable access capital and in- when labor markets are segmented into a formu puts (due for irntane to d nin against sector paying above met-wages with those women etrepreneurs In inform and fom al uable to obtain jobs resortg to informal emcredit markets). Adjsbtm will awavate or ployment. This dualism in the employment Improve these constraints depedi on the pol- srcue can be eacerbaed when adjus*tm icypacage. causes a formal labor sdak-out leadng to the 4.90 Similar procedures can be applied to sweing of the infornal sctor, and a further household enterprises in the saw activity, but in depression of wage-earnings. different regims (stratification I. For example, 4.94 At the margin, the returns to an urban and rural households often face different individuars labor in elf-employent and wagefactor-prices (resulting from input subsidies employment shuld be equal Some guidance which favor urban produces, or differential ac- about whether self-employment is voluntary can cess to fomal credit markets). Credit market be obtaied by calculating the return to labor in disordons benefiting urban enterpris may be self-employment usin the esimated coeffient reduced under adjustment (thrugh finanial 11S on the labor vaae in the profit funion above eralization), and their profits can be predicted (Vijverber& 1988:28a).Giventheestimated wagewhn they have lower capitallabor ratos. furdon (3), the predicted wage for someone with 4.91 Regarding (iii) above, the procedures dis- the mean c of the self-emloyed is cussed under (i) can be applied when the sample calculated. If ths wage Is substanaily higher is stratified between poor and non-poor house- than the retun on labor in self-empkoent then holds. One can thereby establish wheer the tis indicates that self-employment may not be a earings of poor and non-poor households differ voluntarychoice. by activities such as nubfcting because of 4.95 Although the nonrm enteprise sector lower rmrns to education, offers scope for employment expansion, some of 4.92 FInaly, profit functons can predict the its parts are 'saturated' with new entnts and effect of projects targeted to improving the as- profit margins are low. Assisting more people sets of household enterprises, as part of strategies into these ocpations may be self-defeating when to raise their benefits from the adjusent proc- retus to labor are much lower han those ess. In many countries the retuns on owned and wage-e Fmpyn Enourging more fonral rented capital are high, reflecting a lack of access private-sector employment may be a better opto capital assets among households. The earn- to ings of many households would increase significandy if their capital were improved. Similarly Produce-household modes the estimated retuns on the different levels of education together with vocational tainig idi The decsions involved in the operation of cate how useful the country's education struc- household enterprises can be examined explicture is to earing a livelihood in houshold enter- itly through the use of producer-household prises. Separate profit functions for the groups models. To date these have been appl;ed in the outlined above provide estimates of the likely analysisoffarmrhousehold.enterprses(seesingh returns on tageted projects; for example, credit et al, 1986). However, they can be applied to schemes for self-emloyed women, inrastruc- householdsinthenon-farmnsector,withoutmajor ture projects of specia iportanc to rural non- changes, provided that one can assume that fann enterpiss, and targeted vocation assis- householdsfaoeexogenous prices deternined by tance to poor households. convlete product and labor markets (this assumptin is easier to make for non-fam enter- Comparing the rets to labor in non-farm pses -at least in urban aweas - than farm self-employme and wage-mploymeat enterprises). In addition to showing how the demand for own-labor adjusts when relative 4.93 An important issue is the degree td which prices change, household modes also derive the participation in rnn-fam sef-employment, es- effect on hired labor demand. When used pelay in the ua ioma sectr, is 'volun- alongside an estimate of labor supply to the sectay (ie. chosen in prefence to wage-employ- tor, the induced change in wage-levels can be ment), or involuntay because Insufficent wage- calcuated in addition to the change in household opportunities awe avilable. The latter can occur profits. 84

93 4.97 Such mnodeing requies a compehensve Notes data set shice the household's decsions concering tme-allotions, cdmmoditydm ds, prod- 1. The 'actlwpopulon' oers people in wrk, looking uct choces, and hnvestments are all jointly deter- for wo* or presaetly l PeruaseniedIn schodo, or who mined. A multi-topic survey of the SDA type have been enroled during the previo 12 mont are usw pernits the deployment ofa wide variety of model ally excuded. Stadad tnaon definteons should be structures. used (se Turvey, 199). 2. Chpter Three discusses the construction of povet Concluding observations lines under the Sl)A projet The main pdicy interest in Frame 4.1 is how the employment status of the poor differs 4.98 Because adjustment causes profound shifts fom the non-poor. In labor arkets, the formuation of employment 3. For exampl% the Ghana tlving Standards Swuvey Abpolidesbecomesmoedifficultat suchmes (van stract repor emplomt daa usingaroups of: 7-14, 15- der Hoeven, 1987). Empoynt targets in na- 19,20-24,2544,45-54,ad over 55 (tepubbc of hana, 1958: ftonal plan may have to be changed substan- 61). tially as the whole economic base of the country 4 The classifications presented here fowow those of alters under adjstment (Richards, 1986). In turn Chapter Nine below. related fields of g nent actvity (for example, S. he data mod be prend by hoshold-income education) are affected. quinties, but the differe in income sources for poor and 4.99 The kind of infoimation that policy-mak- nonoo houshothep issprcpd isue of poicy terest. ers need thus cages co nsiderably under adjust- 6. Alternatively signfawnt a mey wa diffeental benent. lhe policy questions that arise vary con tween regions may exist to compensate for differene In siderably in nature. For examplen Does adjust- costs of iving. In such case the real-wa differenal should ment influence female urban labor-supply? Wil be computed using te region cst of uving series available financal reform affect the eaning of rual non -fom the SDA surveys. arm enprises? Are the employmet prospects 7. An advantage of the SDA data set is hat labor supply of school-leavers improved by policy reforms? and*mmodity dendanbedjclelye Imdyaeted,snceeach Whatare the effects of devaluation on urban ver- arisesout ofajointutilitymaximization. lis reslts omore sus rural inmnmes? Tbhe examination of these and efficdent parameter estmates udes and Walker, 1M92) other policy Isu iraesa ange of analytical a The areta In de in I(illingnworth (1983). See techniques. Analysis must be accurate if it is to llwomorth and Hedman (1986) for feaile labor supply, serve polcy-making. But it must also be timely and Pudney 908 foran advanced treatment given the pace at which policy formulation pro- 9. Obviously, the values of the explanatory variales an ceeds. Inevitably there may be trade-offs be- beanymnber,dqndingonthegivensampl Theprocess tween analytcal rigor and meeting deadlines, undry the Probit technique translates these values Into and analysts may have to present their condlu- probabilties (which hrefoe rag hm 0 to 1), while mainsions with a wider confidence interal than de- taming the property ihat inreases in fhese transformed valsired. The balance to be struc is a difficult one. ues are associated with Increases (or deases) In the de- The simplest techniques may be reommeded pendent variable (in thisc elp). The cumulative probabilon the grounds that they are the quickest to ap- ity on prwkeambe trasfonmtin,snceitrange ply. However, mispecifyingestimated equations is the (0,11 ItevaL wil led to erroneus condlusions and can seri- 10 See Mlilngsworth and Hedman (1986: 182) for the ously misinform polky-making. Researdhers deivaon of themills ratio. therefore need to be fully conversant with the 11. A msre of scoo quality could asho be induded problems encountered in using multivariate among the educaion variables (BDer and Brdsl 1983). techniques. 12. Weuse prof functon instead of production functions because our concern i with eanng ther than output per se,althoughofcowuethe two funcoksarerelatedbyassuminghouhold pofit (see Lau, 1978). 13. It lows Vijverbergs (198 applia of the model to the COte d'lvore, although some additoa variables are included. 85

94 ~~~~ B; Analyzing human resourc effects: Health The polig issues p tes and water. Reducing the paence of these diseases requires Continued Improvements 5.01 Health is an Importn objective of both in control and moirdtg services, immunizahouseholds and govemments in developing tion and sanitation and public health educain. counties. To households in lowincome societies Household purchasing power is low and thus so lving near subsistence levels, small changes in is the ability to purcae privately procured health household budgets or health program provision inputs - most significantly, food. In an uncercan critically affect phsical well-being. In addi- tain environment, one prone to calamitous tion to its direct Importac to individual wel- drought and political upheaval, households unfare, health indiectly affects welfare though its able to smooth nsumpin through insurance influence on the productivity of work and the or borrowing are prone to catastrophic illness effiency of human capital acculation Con- antd hunger in the absence of outside intrvensiderable public resources in low-income coun- tion. tries are devoted to programs which enhance 503 Structural adjustmtmayinfluencehealth health, poving medical and fmiy plaming by altig the expenditure on prgrams that services, bulding antd aintnng sanitation and provide medical and public health sevices(and deanwaterfaclities,mcontrolllngparasiticdiseases the fees charged for these services), by reducing and subsidizing foods. Public health tehnolo- real inomves by altering wages and farm profitagles have dramatically improved life expectancy bility,andbydchangithepricesof nwmretgoods. in developing coutries without the improve- Not all households wil be equay affected by nient in nutrition, hygiene, housing and income structul adjustent. The relative impact of that cotinded with reduction in mortality in Eu- structual adjustment depends on the nature of rope and North America. The World Bank (1980) state intervention prior to adjusment and the estimates that life expectancy in developing timing, magnitude and distribution of policy alcountries in 1970 would have been eight years terations that constitute the adjutment pakage. less than what It was without the contribution of For example, producers of export crops may new and inexpensive public health technologies benefit from devaluation or the dissolution of such as those based on pestcides and vaccina- state marketing boards. Food producers may tions. benefit from the end of poces which artiay 5.02 In Africa, the disease reduction role of kept food prices down. Workers in state subsicentralzed agecis which rely on public expen- dized entprises may become unmployed and diture is probably greater than in any other part the destitute poor may suffet from reduction in of the world because of the continued prevalence transfers from the state, food-for-work and other of odntagious disease and diseases spread by relief progams 87

95 5.04 Much of the remneconomicslhteraturein cussed in the Analysis Plan on household food the area of health in the developing countries has scurity and nutrition In Chapter Seve been very mkwcnic in nature; making use 506 Suctral adjusp_ent will typcally move of newly avwlable micro datases t evaluate the an economy towards fee markets and away from impact of various inevions - pices and quantitative regulation (rationi irport licenspm-grarm-onhealth, and theinfluenceof ealth ing and ptice control) and lare scale intervenof labor productivity, human capital acmmula- tion in narkets. Given the levels of price distortion, fertflity and intwhousehold resource allo- tion prevalent in much of Africa the movements cafton) Fortunately, the methods and results of in relative prices in the aftmath of a structural this new litature are of direct relevance to the adjsentmay be very lare. How households use of SDA household survey data in the analysis alter their consumption pattern in response to of stctural adjustment in Africa. As the lltera- chas in relative prices, and how this change in ture demonstrates, the analysis of these issues is relative prices maps into changes in individual complex both theortcaly and emplrically. Un- health is thro the key empirical question, obseved (by the investigator) deteinants of and one ftat is addressed at length below. Of health specific to individuals (inherent healthi- courwse, not all dhanges in price can be attributed ness, referred to below as the individual health to structural adjustment. Even with a stable polendowment) and to the locale in whidch the indi- icy regime, prices of many goods - cerwinly vidual resides (the health environment) require foods among them - tend to move over time in specifications and stafistical methods that pay response to other shocks such as weather, crop attention to heteogeneity and the bias it can in- failures, war and prices in world markets. The part. Health is difficult to measure and attention lare seasonal variation in prices in many areas must be paid to measurement eror in empirical of Africa is evidence of a lack of storage and work. Of direct concern in the evaluation of tan ainfrastnr TransportcDstsand structural adjustment Is the accurate measure- lags as well as infomational deficiencies are of ment of the effect of health program placeme-t such Imporance in much of Afica that even in a on health outcomes. As discussed below, if the completely laissefaire econowy one might exspatial placement of health progams is respon- pect that the inttmporal variance in domestic sive to the health attributes of tared popula- prices for intemnationally traded agricultural tions not measured by the SDA survey, simple goods would far exceed the variance in intemameasred associations between programs and tonal prices. The implication is that the entire program outcomes will not provide correct esti- chawe in prices in the aftmath of adjustment mates of progam effects. These considerations cannot be attibutable to adjustment. make the sudy of the dermiants of health, 5.07 Thehealtheffectsofpricechangesdepend and consequently the impact of structural adjust- on the price of goods whose prices have been ment on health, less than straightforward. affected, and the household's abflity to substitute given its preferences and the biological health The macro-meso setting te lgy. Since most food price changes are likely to alter the composition and level of nutri In the comon framework of the Analysis ent intake, inaeasing the intake of some nutri- Plans presented in this Volume, households are entswhile reducing consumption of others,comaffected by adjustment policy through two broad plete knowledge of the full matrix of food and conduits, frasruture and markets in which non-food own- and cross-price elasticities is not they btade. Food nutrients are of obvious impor- sufficient to judge the health effects of price tanceas inputs into the production of health, and changes. Indeed, of the nine food prices considtheir prices and distribution have been the tar- ered by Pitt (1983) in Bangladesh, variations in gets of govenment policy in almost every devel- seve had both positive and negative effects on oping country. Food prices are affected by direct the intake of the nine nutrients considered. 2 Of government subsidy and (much less often) tax, eleven food prices considered by Pitt and excdnge rate policy, aid distribution, arketing Rosenzweig (1985), only one food price unambiand other supply restrictions,aridirect rationed guously affected the demand for all nine nutriallocationtoconsumer Alteraionof all of these ents in one directiont There is a growing body of iervi are likely to be part of a compre- evidence that sugts that the poorest househensive s adstmt. This topic I dis- holds are the most price sensitive, and are liy 88

96 to substitute mote than cer households. There South Asia (soseweig and Schultz, 1982). is of course substantiw Itercountry difference in 5.11 Although attention hus maiy focused on substitutio dasdties and the evidence from ndaer inequality in food allocations, f the rela- Aic s particularly sketchy. tioship between heathiness and productivity 5.08 There is some evidence that for the poor differs across occupations and activites, the discanges in income have relatively small effects trttion of actvities acoss hdividuals within on nutrient co ns on. Behrman and Deola- gender dasses shoid also be related to the inlikar (1987) suggest that income elasticties of rhousehold distrbution of foods, preventive demand for poor households in semi-arid India medical care and other health Inputs Pitt et al are very low, implying that reductions in incone (1990), analyzing idividual and household data result in much less than proporionate changes in from. Bangladesh, found thut there is a pecunay the intake of calories. Again, African speific retum to health and effort in the labor market, evidence is sparse and recent evidence from Bra- and that work activity distributions substanly zil (Strauss and Thomas, 1989) suggests that these influence the intr aeholddistribution of food. findings maybe sensitive to pecification. Inany They also found that eery-intensive effort tends case, nutrition is but one set of inputs into health, to reduce health as measured by weight-or-height and substitution between nutrients and non-food and that thee substial calorie rorcement inputs, such as those provided by pubic health for those best able to alter the energy-intensity of measures, nay be small Whle good nutrition effort - adult males (aged 12 and above) and may somewhat reduce the impact of malaria, di- male and female children (aged 6 to 12). arrhea and other debilitating diseases, it is apt to 5.12 Not all of the costs of m health are borne bea poor subsitute for disease control progras by the individual whose health Is temporarily or and medical intervention. permanny Impaired. Within a household, the 5.09 In addition, studies of the determinants of ill health of one person is likely to evoke resource household-level nutrient consumption provide adjustments by otherperso, perhaps differno infomiation on how a change in the aggregate tiated by gender, in the household In which the diet of a household affects the nutritional intake ilness occur Estimates of the direct effects of or status of indiduals. If Interest in aggregate changes in health policy will be underestimates (famiy levl) consumption or overall nutritional of their true Impact to the extent that these re- "availabiity"inlowincomehouseholdsismainly source adjustmes reduce schooling or labor derivedfromcoerneaboutthenutritionorhealth market exprence of other household membes status of menbers of such households, under- For example, Pitt and Rosen g (1990) found staning how household aggregates map Into that edstig sex-based dif ences In the division well-being and health of individuals, and how of time between household care, schooling and the household distributes its resources among labor forme activities in Indonesia are worsened membefs, is critcal. Indeed, there is evidence among teagers where child morbidity is at a that the intra-househol distribution of resources higher level In particular, it was found that teenin not equal across family members (Behrman, aged daughters were significantly more likely to 1988a; ehumn, 1988b; Pitt and Rosenzwei 1990 I hcrease their participatio in household care ac- Pitt et al., 1990, Rosenzweg and Schultz, 1982). tivities, to decrease their participation in market 5.10 COne aspect of the distribution of foods in activities and to drop out of school compared to low-income settings that has caught the attention teen-aged sns in response to Inreases in infant of many social scientists is the disparity in nutri- morbidity. Conidei these third party" costs ents received by women compared to men. One of Il health will reslt in better estimates of the hypothesis that has been advanced is that gen- effects of policy changes that affect health and a der-based inequality reflects disparities in labor grear awareness of the diffentiated effects of market opportunities across men and women in such Interventions. these settings, with the pecuniary returns to a 5.13 Besides prices for market goods, food housewhold from the allocation of food to women prices important among the, changes in the being less than those for mnal Gome evience supply of health services and the cost to users for support this hypothesis is the existence of a these services are the policy intevents most rlationship between sex differm in infant lily to affect health outcomes in the short rn. mortality rates and diffces In labor market Inmost developing countrie user fees for mediparticpation rates betweefn men and women in cal services are very low or zero. lhe largst 89

97 costs of using medical servics are direct travel the model of household behavior, the reader is cost and the opportty cost of time spent in strongly urged to also read thrugh the more travel and queuing. Thus, the narket for medi- fom presentatin cal service is rationed by the tme costs of travel 5.16 The basic conceptual framework Is an inand queuing rather than user fees Sructual tegrated model of a multimmber household In adjustmt may result in user fees sufficiently which both consumption and producto dedhigh to ration the maket resulting in a very sions are made. In this framework health is a different allocation of service prvision across good which is produced by the household for individuals. Unfortunately, very little work has each of its individual nembers using nutrients been done on how the demand for medical ser- derived from foods, as wel as other health invices responds to user fees, travel and queuing puts. For the discussion to fllow, t is useful to time, quatity of service, informational constraints disaggregate health inputs into those which are and thepricesofothergoods. Anexception isthe individual-specific, such as food consumption or work of Gertler and van der Gaag (1990) who vaccinations, those that are shared by all mnestmate own price and income elasticities of de- bers of the household such as water quality, sanimand for medical services in the Cote d'ivoire tation facilities and housing quality, and those and Peru They find that the demand for medical thatare shared byall individualsand households care is price sensitive, that the poor are more intheregioninwhichthehouseholdresidessuch price sensitive than the rih and that child care is as temperature, humidity, altitude and propenmore price elastic than adult care. sity to parasitic infestaon. It is further usefu to 5.14 Stucturaladjustment mayinvolvereduc- divide all health inputs into those which are tions in government expenditure. Cearly, re- measured and observed, and those which are not ductions in expenditure on health and health- known by the investigator. IfHone were to statitirelated programs are of greatest concern in this caly esdmate the health production function uscmntexts To the extent that public health provi- ing regression teniques having only observed slon is funded or directly provided by donor health inputs as regressors, the regression residagencies, focusing on governmnt budgets may ual would contain the effects of unn_me red provide a ndsleading index of total govrnment (unobserved by the investigator) health inputs -xpenditure on health progams. Reductions in onhealth. Thisresidualmayalsoontainpuly healthprogramsmaybeacaonpniedbyrenewed random component - random in the sense that efforts to taget programs to those whose health it is not known or predicted by the household or is thought to be at greatest risk, or by increases in individual. user fees for non-poor user Efforts at targeifng 5.17 he edolo govenn the production can be ehanced by microecoieric investiga- of health- the health production function -is a tion of the demand for health and related serv- purely boogical technology and not an economic Iees and the efficacy of service provision in aug- behavior. The household's choice of health input ment health. quantities given this technology is, on the other hand, behavioral rather than technological. This Meso-micro analysis biological technology (production function) may differ across types of individuals defined by age, A mode of housdwod bdvir sex and pregnancy or lactation stat. Even hidividuals of the same "typew(age and sex) receiv In this section a model of household be- ing identical sets of health inputs may have difhavior is outlined to depict the major linkages feringhealthstatus' becauseofdifeeintheir between sets of policy variables, biological fac- heaith endowment - the innate healthiness of an tors and the health of individual household individual inclusive of genetic endowments but members. Health status is viewed as purposeful exluding the innate healthiness of the region of behavior conditioned by costs and resource residence - which affects health but cannot be availability, tastes, and stochastic and non-sto- influenced by the household. The health prochastic environment factors. The model's pre- duction function (Box 5.1) thus depictshow health sentationbelowisnon-fmorml. A fom-al economic is affected by food consumption, non-food health presentationof the same idetsispresentedinthe inputs and factdors beyond the household's or accompanying boxes. Although the non-formal individual's contrl, the health environment. presntation conveys the basic implications of 5.18 The household mmizes a utility func- 90

98 lhx 5.1 The healtb production fu=tion Fomly te healthprouton fuchn Is: reorug H*- 11, r. 71 el, 0A i.1... T (1) e ale many suc regoattributes, most of wh are nivlduallyunueasa whch aresummazedasa single "real hedth effect (). Also affectig heaflh b wtere Ji is the hellth of indhivual I (denoted by he the kadlth adummns - the inte heah1ess of an superscipt l,usingm dentaderived from foods(), set indidua Ichuhive of genetic endowments mt u exudig of nofood health inputs (Y) whkh do not provide utuity nate healthinme of te region of ridence whdch Is ireedly Onredical sec suc as inoculatns and other In ated In IL- which affect helth but cnnot be inprevetve care, public health fadties), a set of household fluenced by te household. *publ reources (G) which affecthe health of each famlly member jointly (water quaity, sanitation fadlities, l he functms ) may differ ass person-types (de- fined by a8e and sex, for example), thus permitting the housin& qualty), the time of certain housewhd members e ffects of health inputs on halt to differ across types of (l devoted to such health-related activities as food and individua. An altentive specificatin to the T sparate waer peparaton and cl cre (mothers time), and the hedth production hmdos of aton (1) would be to (posibly) deletrous effects on health of effort expended have a common functional form for the health teholo workg (e) Attibutes of the regin in which the house- - the supersacript on the functonal He) can then be hold reddes, such as temperahte, humidity, altitude and dropped -but add to the Ist of health inpub a vector of propenity toparsitcinfestation, abo affect health but we individual caderics - age, sex, and other healthlocade-speic rather than household-specific as are the relevat peonal chartistic tion which includes as arguments the health of increased health results in increased income the individual menbers as well as foods. It is through the productivity effect In addition, to sensible to include both health and food con- theextent that thepecmiaryretus to increased sumption directly into the utility fuion since health differ across meber types (age and sex), health Is desale in itsef and because foods are the household will have an inoentive to allocate consumed for reasons l other tn their nutritional more health inputs, such as food, to those indivalue. The utility function also inctudes the lei- viduals whose value of work time is most sensisure of individual household mebers, their tive to increases in health. In the case of a farm cosnptn of non-foods and their work effort, household, the marginal product of work time the latter being a bd (utility decreaing) rather may depend on the level of oer farm inputs than a good (utility augmenting). This formula- used in production. Marginal products of labor ton is very general in that it permits households for any given labor alocation wil be higher for to assign different wghts to the consumption by households with lawge quantities of land or farm every individual of each good, including health. animals, for example. Thus, the productivity re- Allowing for the possibility of discrimination in tuns to health will depend on the level of all the allocation of resources (including health) fam inputs as well. auvmng membes and types on the basis of sex 520 Households maximize household utility andageispiudentingeneralbutevenmnoresoin subject to thehealth production function,a time *he case of Aftica consuainthat specifies that the total time avail Health may affect the productivity of la- able to an individual in any time period can be bor time and hence income. If markets recogize allocated to work, leisure and the household care and eward health-related productivity ditferen- activities (child care, household chores), and an tlals, then the wage for an individual is a function inome constbaint that Includes the return to of his or her health, the strnuousness (energy household farm activities. The process of conintensity) of effort required by the individual's strained household utility maxmization is foroccupation, and the personal charteristics of malzed in Box 52 (oveieaf). the hdividual, such as scoolng. In farmhousehold, health status may direcfly affect agricul- Reduced-fm demnd equas for halth tural productivity in addition to productivity in the wage labor market. he implication of these 521 The benefit of explicitly setftng out the productivity effects is that they tend to reduce structural relations that underlie houewhld bethe net cost of attain any level of health since havior is that it informs us of the fl set of policy 91

99 Dox 2 Constrained utility amization Fomya ll hobo d,consitingoftldvidualmaxd they ea In the labor maret and In househod fam pro, mizesa ulity function of thefornm ducdto I~~l.,T (2) ~ T u=uu<h, K LI' T, z,z,ei), i 1,1... T /2) T Yf+v+n.zp,5( )+P Ezi+pyt (6) ZW''+VflZPkZC)q z' I, Zy'(6 wkid indudes as rguments the health of the individual membes as well as foods (C.it is sesbe toinlude +AW (V+.1) both health and food consumption diectly into the utility fhmdslnhealteiisdesdrblneifandbemufoods where llis hm prollt ameomnwdfor reasosoththathelwnulina valu T Theuily funidon abo induds the ltisu of inddp x-zwi -ia househl membm CL, their consum Of non-foods Q F p a') and theb work effort (Ar Mhe total tm available to an Individual housxel Wionwprdnd enpisome), nember i (' cn be alotd to maret (wage) work w' Z (the non-food amsumpti good), PQis the price Of flu household cae activities Cl bbisure Cm and, In fam r houseold, to actvties asodated with erop pi frim output Q, Ps tfe podce of frm mel fnputs, and and processingft 1 g lhz~~~~~~~~~~~~p TMu') at efimbudg isif isebdes Ike lhprce all human of food nutition kb cme from foods, the health FI+ 4+f+ Ll 13) pmh fi) etkm)pucinnto uldhavebeenwriae withfoods 'C as inputs instead of the nutrients N. To acmpte the Health msy affect the produdivity of labor time If formulaton with nutrients as ipt, an exp aion conmawket inieooueandrwdhe-lt ivndpo ivy verting nutdents into foods can be added. Howeve, this diffentla then the war for indmdual ib a function of addmanal equation adds nothing to the disuion of this his or her health, thea (energy Intensity) of sdcod effortqedby te Vis 6e), and tte There are many distinct nonoods cosumed by pesond dwa s ( of thhe indidae such s ds so that consumption of nonfoods by indivkdsd') the IndivIdual, such as a, Z, is a vector of goods as are foods C'. Howwer, because of the key role played by food in the production of i el,i) (4) the individua h ele menteof the vectorc'w emad. = ~""'k '~"~ '~' texplicit by the use of sumnntons over vector elements in In faim houshdds, halt sts may directly affect hat follow. Fo notalo simplicty,z'and agricultua productivity In addion toprductvityin the other veor of goods to be intoduced below are not so wage labor mael The nfam prducion functio Is d- treated.,crib,_it y: does not matter for our household modal wheher the UCrIbOdW household utility function esents the pfrence of an Q=Q(P,,XA.KKH,y,u) (5) houwehold acfamu mwobers, fydct or inthe unlifedsetofpreferncsammg result of abargainingprocwheeqisimouput,fisnon-fmfmllybor,,pishfed ess. The list of aguments that appow In the (kedu agricdulunal labor, Xis a vedor of agiultual inputs, A b form)equatio tdbe estimated areunaffected by thederiland under cultivation, K is a vector of capi input, and vation of the hou utiiy functo I, for exampl it Handyw are ve of healtb stau and peonao Chan- is found that mos wages have a bigger effect than.t of famy nembers, r,espetively husbands wage on the alocaion of health soures de- HousewAds I dize theif d uti fu don CZ) voted to daughters, and the reverse true for health subtsttolbehompo& mhek ),thetmhneudt sallocatns to sons, one coanot daim that tlds unmbigiconraint (3), the wage funton (4), the farm production ously etabis e vare o bagon mode in funcion (5) and a (fl) Inwome cnsbalnt that resticts wi hghe wages give one more baganing power and tmm to spend no more on goods and.'eath inputs than lop l wih d moes variables that may affet headth and by what houseold's demand for health for househld nm ism they work. In partcular, this house- member of any type (age/sex isa function of all hold maxizdntion problem leads to a set of de- the exogenous varlables: the prices of all goods mand equations for al household choices as func- (inluding time) In the utility function, the health tionsof all the exdgmog (non-household-dice) production function and the farm production variables - known as the rduced-f6nm demand function; the health endowment or innate equation. In its most general form, the heltbinessof eachmemberof the household; the 92

100 region-specific health envlronment the gender vdaracteristi of the cultivator (u) and othier (exog us) personal cha cs of can au be dropped from the demand equatin (8) every household memberf and the quantitis of and replcd simply by meured farm profixt land, capital and unobsered factors that affect (n). The reasonableness of the sepaity asfarm output The reduced-form demand equa- sumpdon will certay vary across rion in tion for health is presented in Box 53. Afilm Separablity seems reasonable for cash 5.22 The following comments on the health re- crop farmeus, such as cocoa or tobaccocultivators, duced-forindemandequato arein order. First, but less so for food crop cultivators for whom by including the acth quantities of land and marketsalesandpurcasesofagriculturaloutput productive capital rather than prices, we are as- or agricultural wage labor are rare (subsstence suning that these quantities can be considered fwarers). 10 fixed, in that they are not choice variables (at 524 Fifth, of very direct concern, the prices of least in the medium-tenn) to the houshold. Sa- the non-food health inputs (PY) are typically not ond, in practice the housing quality goods (the G available, or If they are available, are not ecogoods) can also be treated as fixed to fte house- nomnally relevant in that they do not represent hold so that their (typically unmeasured) pnces true (private) opportunity costs For example, (P) can be replaced with a set of variables which even if inoculations are fee" in that no fee is measre the quantity of these goods consmned collected fron the inoculant at the time of inocuby the household. 7tird, the individual health htion, the oortuity cost to the household of endowments ji9, the regional health environ- aninoculationmaynotbezeo. Itispossiblethat ment (?), and unobserved factors that affect farm not everyone who wanxt an inoculation gets one. productivity (the cultivation endowment u) are This may be beause there is an insufident supunobseved by the researwh, but presumed ply (excess demand exists). Also, the true "pric known to some extent by the household. As they of an inoculation includes important pecuniay are urknown to us, they will form a part of the and nonpeuniary costs to the household in adregression residual when thedemandequation is dition to any inoculation charge. These costs estimated. may indude the cost in terms of time and monwey 523 Fourth, If markets are sufficiently well of getting to an inoculation center, a cost which functioning, the farm household can be thought may not be inconsequential many areas of of as solving its fanm production and utility Africa. In addition, there is the 'cost' of acquirma~dizationd problemsreursively(inseqwce). ing informatio An individual who is not well It will first act to maimize farm profits in its infomed about the benefits of inoculations or choice of farm inputs, and then maximize utility how to go about gettng them will behave lhke given these profits. This recursive structure - one who confronts a high prie - he or she wil known as sepabity - means that the prices of not get inoculated even if no fee is levied. Similar farm output (Pd, the hired labor wage (W'), reasonlngappliestoother public health progams prices of naterial inputs (such as ferilizer) P) and interventions such as pubuc health clnic, quantities of land and capital (A and K}, and the fainly planning clnics and the lke. Other types Box 53 The reduced-form demand equation for health Fomaly, the rdued fomli cnipl housing and water quahty, anin fadilid H~~~ - wr ~~~~~ ~~of P heboseok, thebhealth endowment onamteehea1tbi DR'(P1---- w **W W P. V DH nessofea&,of hetmbesof thehouws&aldilthrough P!,+ P, WI',WDl..., D T, A, r u) (8) IJ), thereg-spedftchealt hvln toc thepudicof the rm) output poduced by the h_oshold (P.), the To th reducd-form demand equaton, the demand for pioe of all pudaed inputs into produdng ts good halth of pfmon tpi depends on the pri of all K foot weratepadhtred workers), thegenderand goosprices(p, through IV, the wages (time pics) of al oie (oo) peroa chamcteics of every ous-- Th mbesm (WI though Wrl), the priom of non- hold member (D trough D T ), the quantftles of land food goods (P 2 ), the pric of healt inputs whkh do not avalabl for cultvatio (A), productive capta (10, and provide utiy Sdy (Pr the Il lnputbs)ahe piat {the unobved dartertks of fte lad and ctvator of imuseod EpubIe resources whih affect health (fow which affect productivity (u). 93

101 of pubic health initiatives by their nature do not price of a food on thedemand forhealth. Asboth involveafee-for-servicebecauseof thepublicgoods food and health are goods In fte utility funcdon, natue of the services provided - mlaria con- the mgnitude of thep ice effect depends on the trol and the spraying or draining of the breeding ext to which they are substitutes or compleplaces of water-bon parasites and other disease ments in prefern. But as food is also a source canrying organsm ane examples. If it is true (as of nutrients which are used to produce health, is typically assumed by most researchers) that the increased price of food will also increase the health program placement is exogenous in the price (or cost) of achieving any level of health. estimation of individual-level health reduced- The size of the increase in this implicit (shadow) form demand equation, then the availability of priceofhealthdependsonthenutrientcontentof (or traveling distance to) public health programs the foods, and the marginal rates of substitution and failities can be used as regressors i lieu of in health production among all health inputs. prices. Unfortunately, considering the possibil- Thus, fte total (reduced-form) effect of a food ity that programs are unlikely to be placed ran- price increase on the demand for health includes donmly throughout a country but rather placed by influences from both prefernces and the biologigovernment in response to observed regional cal health production function. Because of this health deficiendes, makes the comrct statistical complication, the usual strategy is simply to asmeasurement of health program effects on health sume a linear functional form for all the reducedproblematical Adjstment policies often result form demand equation. Furthemnore, the uhobin significant chmages in public health prgam served health endowments s! are assumed to enbudgets, so that it is quite important that this ter additively. We can think of the residual (t') research obtain an accurate measure of the influ- of the linear reduced demand equation for health ence of public health program availability and ascontainingthe own- ( and coss-effects )of placement on individual health. Alternative health endowvments on the health of individual i, methodsof estimating theeffectsof public health in addition to all other sources of unexplained program placement on individual health will be variation (r,): addressed at greater length below. For now, we wi treat public health progam placemnt as p=;o + &t +,q" (9) exgenous and thus replace the price of health I" programs (P.) by measu-res of availabity or dis- 526 With the above points in mind, the estitance (Y). nable reduced-forn demand for health equation 525 Sixth, It is difficult to determine theoreti- can then be written (supresn the superscript i): cahy apprpriate fnctional forms for the reduced form demand equation (DI'(o)). By choosing a T T functional formfortheutilityfunction(equation H=o,+EaOP P+1 fwi+zt14d1 (10) (2))- such as the Stone-Geary utility function t1 I i l that underlies the simple Linear Expenditure + ORG + inpy + 4 II + pv V + System (les) or a quadratic utility function - and a functional form for the health production where the Greek letters in equation (10)(except 1I function, one can not (in general) analytically and e) refer to parameters to be estmted, the solve for the health reduced-form demand equa- superscript H on these parmeters refers to the tion. The demand equations are not LES (even if fact that they are the paranms of the health the utility function is of the LES (Stone-Geary)) reduced-form rather than the reduced-form of form because the maximization problem of a another household choice varle. Also note household that must produce one or more of the that this version assumes that the health reducedgoods it consumes (health) differs from the usual form demand equations for evety nmmber type problem of utility maximization subject only to a are identical except for the sets of d 1,...dr personal budget constrant Treating health as a utility characteristic variables (age, sex). Altematively augmentng good produced in the household one could estimate separate health reduced-form implies that its demand equation is a combina- demand equations such as (10) for sets of indition of the demand for health from the utility viduals defined by age and sex. function and also the supply of health from the 527 Reduced-form demand equations for all health production function. For example, con- contemreo detminedhouseholdchoice sider the impact of a compensated increase in the variables (necessarfly) have the same set of left 94

102 band side va;iables (regressors) as the health ye- whethe or not the tespondent stayed in a hospiduced-form. There may be some interest in td or health center and for how nny nights. stdyg the effcts of structural adsust on Clearly, this measure of headth is likely to be the demand for health inputs as well ason health strong related to the sodoeconomic status of outcomes. The analysis of the demand for food thehousehold conditioningonactual ealthiness. and food nutfients (which are important health 529 There is soe evidence that the other set inputs) is discussed in Chapter Seven. For ex- of health data available from the SDA survey, ample; the reduced-form demand equations for miscarges, stibirths and the death of young non-food health inputs Y' are (supressing the su- infants, tend to be under-reported by households perscript i): in similar surveys. Furthermore, it is not likely X.rWT T that the probability of not reporting such an event +Y x P+ + iw'*ymdi + (11) whenltinfacttranspiredisu reuatedtohouse- -1u J-1 J- hold attributes and the economic environment, +80YG+IOYP 7 +Vfl +pw+e Y 530 As the pr redanthpotrcdataare available for persons aged five and under, the Variables in the reduced-form health mnesurement error in self-reported inducators of health is of consequence in analyzing the effects 5.28 MEAs1 of Ho ALTH. The SDA Integrated of strucural adjustment on the health of a major- Survey provides four types of data on health ity of respondents. It would be a mistake to status. the self-reporting of respondents as to ignore this sub-population or to think that selfwhether or not ty had suffered an illness, the reported health indicators do not contain useful effect of any illns on the perfornipnce of cus- infonnation on health in spite of the measuretonmy activiies, the anthrpometric measure- ment problem Although it is likely that the rement of height and weight of children five years duced-form coefficients associated with age, and under, and information on miscarriages, education, wealth and wage rates of household stlbirfts and indant deaths for pregnances within memers will be estimated with bias In an equathe past twelve months. Anthropometric meas- tion having self-reported heolth status as a deurmnt is less liklty to suffer from measure- pendent vaiable, it is likely that nuny of the ment error than the two self-reported health sta- regssr in the health reduced-form are not tus mures. Of partcul concern for the self- correlated with self-reporting n tmuremet error reported indicators of health status is the possi- and consistent estimates of their effects on health bility that the propensity to consider one's state canbe obtained. In particular, all the market-and of health as being ill" versus 'not illr may be region-specficvauiablessuchaspricesandhetlth correlated with individual and household char- program availability are unlikely to be highly actestics such as education, age and (full) in- correlated with self-reporting error or with the come. Pitt and Rosenzweig (1985) used self-re- individual- or household-specific variables corported ilness (whether the respondent was sick related with self-reporting measurement error. in the prior week and whether he or she was sick Ihus, it is still likely then that the effects of the in bed) as the dependent variable in estmating exogenu deteminants of health most directly separatereduced-formhealth equations for2,347 affected by structural adjustment will be reafaormhouseholdheadsandwivesfromlndonesua sonably wel nmeured. Nevertess, anthro- Theyiterpreted thepositiverelationshipbetween pometric nmsurement of health status, in parrespondents' schooling (in years) and the proba- ticular weight and height, are stil to be greatly bility of reported illness as resulfing from sys- preferred inananalysisof the economi detenritematic self-reptng measurement biasu" The nants of health and the impact of sructural adother self-reported indicator of health status, justnent on health. Enumerators will need difwher ilness has altered the usual activities of ferent weighing instruments for adults than chilthe respondent, lsalso likely to suffer from meas- dren and some additional traiing in order to uremnent error which is correlated with some ac- weigh and measure adults. tual deterninants of (unobserved) true health 531 A more satisfactory measure of self-re -the opportwdty cost of time, and the maginal ported health status would make use of a list of utiity of health and income. This problem is reported health symptms, such as fever, diarprobaby more acute for another measre of health rhes, skin outbreak and the like. In areas where status available in the SDA questionnaire: certain types of disease are known to be wide-

103 spread, sympoms assodated with these diseas the aveme unit price in a locality (village) for can be Itemized in the questionnaire. s type of each food category. measure my actually be more Informadve in 534 Wgs. Wage rates can also be derived as some repect than weigt and height although "undt value from the houseld qusbonne more likely to suffer om self-reportng bias. or f*m a commity survey, the latter being prefed. Ifthe househod quesmktiae Ib used 5.32 REDuRw M o*adn=amsjophnml The I is rneway to calcuate the average wage for regressorof the health reduced-form suggested each gender by swhooln vel and season. A by the conceptual model descbed above and more sophstcated approach would estimate seavailable in the SDA household survey include: lectivity-corrected wage equationsby gender and * Prkc (P ): prices of foods (see also commu- use these wage equations to predict wage rates. nity level data) In most Africn contexts there would likely be * Wages (W 9: predicted wage rates of head of lttle retuto thekeffort required. household and spouse 535 Houseold public goods. Most of these are * Indiuidua attributes (dl): gender and possibly categorical variables in the SDA questionnaire. age, ethnicity or religion of respondent Depeding on the distribution of outcomes and * Housedold 'Public' goods (G): sources of snple size, it may be necessay to aggre drinking water for the household, nature of rub- somcategoral outcomes. Forexample, twelve bish disposal, type of toiet, physical attributes of choices for source of drinking water would rehousehold dwelling (typeof dwelling, numberof quite 11 binary (0-1) vriables Some of these rooms, whether dwelling shared, source of light- may be empty (never chosen) in the data or be in& main fuel use for coodng, constuction ma- chosen by such a smal number of households terials used in outside wall, flooring and root), that reasonably predse estimation of thdr health ag and education of head of household and effects is impossible. Investigators should then head's spouse usetherjudgementtoag te categoies. Some * Prices ofnonfod health puts (P.): measure of other regressors are probablyof only second order availability and accessiblity of public health and impaasdetedinantsof health orae highly fertiitycontrol program (fromcommunity-level coreated with each other. In partcular, not data) indluding as regressors the full set of variables * Nonearnings ome,faim prot and wealth (II describing the contrtion of the dweing will and V): measures of wealth Induding land likluy not have a consequential impact on th ownership,value of liveswok, value of agricul- esimated paranoete of interesl tural equipment, assets of non-farm enterprises, 5.36 Pris of non-food hadth inpu. As noted lonearnings income and farm profitsl above, meaures of the avallability and/or dis The following comments are in order. tance to providers of health and related services Prces should preferably be derived from the obtined by the commurdty survey wil act as conunuity-level suvey rather than by calculat- poxies for the pries of these services. Investigaing unit values by dividing the reported value of tors should be careful not to use as regressors the onsmption or expenditure by a household by infirmation on actual use of health services and the physical quantity. Calculating household- the cost of travel to the service location provided spedflc prices by the unit value method imarts by the husehold questionnaire. The primary bias if individual food categries consist of items problem is that use of health services is a housew of valying quality. Higher quality food has a hold choice, a dependent variable rather than an higher unit price than lower quality food, and egno independet variable. There is obvisince taste for quality is (ikely) a nonnal good, ous"simultaneousequationsbias"rwtingfrom price and income effects are codounded." This using as regressore the survey infomation on quality bias problem is more likly to be a prob- wheiher an individual had consuted a health lem for heterogenous goods such as "beef' and practitionerorvisitedahealthcenterorconsulted "fish & shellfish" than for more (locally) ho- anyone about an ilness. One is more liely to mogenous goods like kerosene or salt. If inde- have consulted a health practitioner if one has pendent measres of prices for constant quality been ill so we would expect that visits to a health foods are not available from the community practitioner and one's healt woud be inversely questonnaie, much of the quality bias problem correlated but would certainly not want to draw of using unit prices can be elimnated by using the condusion that visits to a health practitioner 96

104 cause poor health The same problem exiss even coutrol for an Imporant convte heth of status, for the use of pretive medical cue - he less such as age, while tabulating ainst commudty healthy a person's Inet healhns (the kss to importat publc health program (se their helth endowment), the morelikely that forexample Frme 5.1). person will seek out prvtive nmial cane, all 5A0 Condioordg on age as in Frame 5.1 may else being the same. be ufu to undov the efects of heath Pr- 537 Nonnngs inomep awixdwndwui grans that primay affect the health (as meas- ItwouldbeusefultDcreateavarv bleforthetot ured by weight-fr-height) of Wants (age less value of assets from the data provided in the than 2 yeas) er thn older dren Cross- SDA survey. Th exogeneity of all sources of tabuions such as these thus hlup rse,es nonearmin income is problatic. Transfers uncover vstylzed facts" about the association that are not "regur may in fact be a reflecti betwe health interventions and health outof the ill health of household mems or ote comes However, coss-tabulations necessgily household outomes that reflect household involve orny a subset of the variables the prior choimes. One cnmake the case tat even rel- analysis has indicated are behavioral deterndlaer trfers repreett past consumption, in- nants of health. Crsstabulatons thus only provestnent and nigration decdons of households vide us with parti assodations and may be misthatarecoelatedwithunobservdded m leading. Forexample,a variablechosenforcossof health, hence resultg in biased coefficents. tabulation with a health outcome may be highly To the extent that none s Income is unm- coreated with a variable not In the same crossportant or zero for most huseholds this Issue tabulation, and may thus be pikidng up the effect may be mool However, in some African coun- of theomittedvariableandnotjustitsowneffecl tries, remittances from household mems who Past reseah with this type of data has indicated work and reside away from their iumediate that ths "omitted variable" bias can be severe. fandly members may be an important souwe of For this reaso a regresson that contols for all non-earnings ncme to the ample household. If reduced form determants of health is the prethe remitters reside outside the sampled house- ferre method of anaysis holds village or town, it may be useful to treat this as eogmous nonearnings income - in- 5A1 RRGssON ANALm Estimation of the clided in the reduced forml helth reduced4rms using an nic _msure of health as dependt variable is sim- Eshntim mehos ply accomplished by ordiay east squares regressionl J3stikmtinoftheiealthreduced-forms 538 TABUR ANALws The first step in the that uwe bnay indicator of health such as selfanalysis of data should always be simple tabula reported lness should use probitorloglt esmanalysis. Creating some simple tables is a quick tion meth Tse procedes are found in and easy way of gettg a "feelfor the dataand many popular eoometric software packages. wm ilndicate whether the range and variation in There are some problem with the reported t- variables of interest are sumfficient to use in re- statstics wen ordinary least squares regression gression anlysis and if outliers or coing erros (or proubit or logit) are applied to these data. A ae present The fist tabletobe produced shoud more tehnical disussionof this issueispresnted be a table of means, standrd deviations and in Box 5.4 (overleao. The reader is stmngly enmlndmaandmximaforallthevaiablesdescribed couraged to read this material as it suggests a above FoBlowing this table of descriptive statis- F W of children by tics, a set of cn should be under- a p4 s avdpa f cy taken to exaine patterns in the data In partkcu- age and pam availabt lar, health outomes suld be tabulated against gender and age, region of residence (or ethnic- Haf pwpum H,M pmp,, uo ity), mea_s of wealth and Income, education inammwit pram in wumnu of household head and spouse, and availability of Important publc healt and fimily plarnnng A 2ssand2 program and mdcl provides A more so- As emd5 4 psicated cross-abulation dight attempt to 97

105 Box 5A Tecdcal issues In _gsion analyis Whdie ordinary least squs regressin provides mls- random efft Ithe cipoet of dke reiduals that are ttup.raeteaestimatsunderforlygenexaddrcw mstances, - no bdep t as obsvatons in t case endov with housebold survey data I may tend to give estmtes mentsl re not a ated vwit the regraorn whih Is te of te sandard ems of parameters that are assmm mnaeaboutresldualsinaflreduced4orm etlbiasedownwards- that Is, the reported t-values wl be man). Many ammon statistical paccae, bhave random based upwards This bias occws when the re ion residul are not h epdenl Thi lack of indeptdene effects estimation built-in. Rxed effects esiumati teatb te commonality among sample o k essentiay can bethe reut of (in order of magltudek (O) An Individual appes more than once the estlas a fixed parameter. For example, if an Individual i smpled at me than one poiatin tinme, al ime iwnariat mution smple becausehecmshewas sampled at more than (observd and unobsrv) chaactesatl can be capted onesurveyrourd and multplesurveywrus werepooled by a (biay) dtummy variable specc o the IndividuaL for estmaion. As th regesi reddual Indudes time Par even smul samples, this resut in a relatvey lage Invaamt Individual hath ernowmb, the redduab for mnmber of extra ters that need to beestimated. Th an iualarelr k w tobecordateda oss timepaod problm of too many parame i elimiated by differ- (U) More than ain Individual in a household appeas endngaesstnimepeuods-aprocedure which results in in the esmation sample (even in a single crowsection). the same pwaameter estimates as explidtly indudng Members of a houswdod are not anly lkely to have indi- dummyvaribles (E1minatingunobservablescoinmonto vidual endowmeb n ) that are orrvated, in pat rpeaenting unobserved househob dspecfic determinants of al hosold members by houehdd fixed efects entaib dfrendng the rersso for each individual from the heth, but also they shwre a common health envronmet average for all household mewmbe An example of this (X), many of the attributes of which are ubsoved by the aprach to studying the hntra-househd of 1nvIlSgator. health and food is Pitt et al s990. A ssequence of (I) Mowethanoneindividualfromareglonappearsin hou_hold fixed effecs is that the effecb of householdthe emati samples Again, their regression reul spedfic detminants of health [incuding the houed wil be corelated bee of hiwer common incompletely pubc goods Gl cannot be Identified that the fixed obseved imear)lrgoahealthe ment0). Note effedsmodelpreustheesmatioftheeffectsoftime that in the case of mtiple survey rnds, indivua invariant fat (such as gender and many houeold meaured in a common srvey round wil wil hae come- caeristics such as drinking water soure) nhealth. lated redduals if there timepefic factors such as When estimating health reduced-fo_m equa_io for season, prevaence of cotagious disese, or other naturl persons over the age of five, sef-reported 11ess must be phtwh wbhic Importantly affec medhealhel asitent esimate of thiestand err- of coeffila used as the dependent variable- Secause Ihe dependet varible isdkchotmonu, logit or probit model shoud be dents can be obtaned by estmating a random or ued atede Unfortmately theproblem of celatd eos effects nodel Randome dfectb estimation is a fonm of dsedaboveismeytobeapblemthatistoodlfbult geralizedlstsqutatmamkesuseothefinoati todeal witho nputonalywhe thedependentvariable that the residuab have a corraion stucture (random iso continuous. effects estimation also requir the assumption that the variety of ecoronietric specfications that are dependent variable and all (continuous and usefl and not difficult to implement nonzero) regressors, then the coeffidents have 5A2 The interpretation of th regressin coffi- the familiar elasticityintepetation In the case dentsisquite straihtforwdin a reduced form. of prices and program availabilities, these elas- In the case of linear regresion (without interac- ticities are the direct effect of a percetage intions), the coefficents are the effects of a unit crease in eac price and program on the health chag in eah regressor on the dependent vari- outcome. able. For exanple, the coefficient on the price of SA3 lnepretationofceffiientssonlydsghtly milk in a weight-for-height reduced fonn is the more complicated in the case of a probit or logot effect of a unit change in the price of ndlk on the regrssio Partial derivatives, which have the weig-for-height of a child. U nmik prices are in same intpetation as a Hlnear regression coefficurmncy units per liter, and if it is known that dent (above), and elastictes are often output structua adustment resulted in an Increase in from computer packages that estimate these the price of milk of 10 curen units per liter, models. then the effect of this inrease in price on weightfor-height is simply the regression coefficient Other prblems tines 10. If natual logaritlums are taken of the 98

106 5M44 H.miH PRoRAi waceiant. As noted tive coefficint that health prg_ram investments above, tee remains the issue of whethe the moad ill health rather health programs were reduced form estimation strategies descibed placed where health endownents were lowest above aourately nmsure the effect of public The result of this 'reverse causation is to underprograms on health outomes. A fundamental estimate the effectsof health program on health problem in progrmn evahation is that the o- out s when the latter is regressed on the fortion of progras and the timing of progrm ird- mer. tiatives - program placement - is not liely to 57 Data on the spatia distribution of health be random to the exten that govmrment deci- program and outcoms at more than one point sion rules are responsive to un_msured attrib- in time, expected to be available in the SDA surutes (regnal health endowments) of tageted veys, can be used to conistetly identify propopubtions. Simple measred associations be- gram effects. Fixed effects estmaion (see above), tweenhealthprogamnisandhealthoutcmnay which estimates how chnges in local progms not provide corect estimtes of prom effects affect chmges In the health of the local population and thus mis-esima the effects of structrl are free frm the containation of arel heteroadjustment on health. geiteity bias. Note that if the points in tinm are 5AS Inanycountry,atapointintime,progam too -dose", it wil be difficult to get good esliefforts vary widely across areas, even if the pro- mates of progam effects since health pwgram grams are funded and controlled by the central availability will have hardly changed at all. governwmt. Access to public health dircs, hos- However, if a strcural adjustmnt program pitals, umanity inics and fertilty contol pro- impotantly affects the ability of the govenuent grams varies conid y from region to region to maintain health pogram, sgaificant chnges within a country Given the limited resource in health progm availability will be found in apacities of the central public agency, program the aftemath of strucual adjustmen. If te alloations must be rationed. The plaement of is in fact little change In health program availaprograms is thus likely to depend on the ex- bility after strucbt adjusftent, then there is pected locao-arspecific returns to the progra, lkly to be lttle change in health outcomes as a which will vry across areas according to, among consequence of changes in health program other attributes, their physical and denwgraphic availabflity, so that the measuremett of health characteistis (endowmenit). If progam place- progm effects becomes unimportant mentisattat ive to l -secific endownm ents, and such endowments influence outcomes of in FuNcnoNAL orm AND sanfle UmSAcGRAterest to policy-makers, it is important in pro- lnon. The reduced form demand equations for gram evaluation to have infrmation on endow- health and health inputs preted in equations ments. It is inevitable, however, that not au ex- (10) and (11) are written as linear forms simply ogenous locaional crteristcs awe measured because of the computational simplicity of workor are measurable. This is surely the case in the ing with linear-in-the-parameters esmating Afrian countries paricipating in the SPA Pro- equations. If degrees of freedom are adequate, gram invetigators uay wish to alow for interactions'r 5.46 For example, consider the case of a coun- aitong variables where it seems such interactry which has certan areas that have environ- tions are justified by some priorinormtiaon about mental features favorable for the propagation of behavior. For exumple, it would be sensible to geograpallyconcentrateddebilitatingdiseases interact prgram variables with mothers' educasuch as malaria, river blindness (o ras), tion if there is reason to belive that the effect of trpaomasis (sleepng sickness) or schis- ceain progam ievetions on the health of tosomasis (bilharzia). The cental goverrment, cldren depends on mother education. There knowing this, places a d oponate level of maybe also be policy reasons to esdimte models health programs in these regions relative to re- thut permdt an analysis of the incidence of stucgions in which these diseas are less frequent. turd adjsbent among variouslydefined groups 1he comrelatin between frequency of illness or If there is some reason to bve that the althiweight-for-height and heath program availabil- ness of these groups (or demand for lth initybased uponhusehoiddatathatrelieson spa- puts) may respond diffely to changes in retialvariationmaythen bepostive. ltisclearthat duced-form determinants, the investigator nmy it would be inornect to condude from this nega- either introduce a variable which defines groups 99

107 (rumi/urban loctn, rgio) with reducedfiorm ouy lnadequate in eauatigs effectiveness regressor or break the sanpe and estimate and the effect of chages in progrm availty sepaate kud egchgoup of interest, on bealth.' Statisdcal tests (Chow tst, F tes) cn be used 5.50 Unfortunately, conect for the selective to determine if such Ition lbag or interac- effecs of nmrtality I difficult. It is necessary tion I statisticay vaid. (but t suffici) to know the effects o programs and prics on the rate of survival a the 5.49 T1 PR op OF MOAfLY AMD PSfnw se- average healthof survliors. Knwing the former uecun. Drawlnginferes on te effect of health requims Infomatio n onmortality durig a (com- Intervenions (Inluding food prices) on children on) referene peiod. The curent SDA pintofrom stdies which are based soely on samples type survey questionie only provides this inof survmg chldrenmay lead to an incoet as- fomation for infants bon in the twelve months sessent of progam efecs if the effects of pro- prior to the urve date. Considering hat nwrgram on survival and fertiity probabilities are talityratesarehighestinthe flityearof l, this ignored. In many African countries, inhfant mor- is still useul Information. It would be better to tality rates are very high and health program record hifomaton on all child death in each which affect cld health (as meaured by height wonmn's reoductie l d the year, sex and and weight, for example) also affect the probabil- birth order of all live birts and whther they ity of survival for children ever-born It is not resuted in surviving infants (at the time of the 3en.outoconsidertheimpactofhealthprognams survey) or deaths (and when death occred). and goods pries on child mortaity in addition Ssticalmetiodsexdt (andareavailablein som to their impact on health. Esmating the re- canned" statistical pakage) that measre and duced-form detminants of cild health is fur coect for the sdecdon bias Impated bynonranther complicat because chmges in morality dommnorlity. Asimpletoimpleenttwo-sage induced by changes in progams which affect method first estimates the reduced-form deterhealth (due peihaps to strucal adjustmn may minants of mortality using the same set of regresalter the average health of the surviving popula- sors as equation (1O)." This estimtion stage is lion by causing those with higher inherent valuable by itself since infant mortality is an improbailities of ilness (those with low health en- portant health outcom and the effects of str-m dowments 'and A) to not survive. It nught be tural adjustn on mortalty as well as anhoexpected that the LHkedhood of infant death will pometric measures of health on living children be inportanly related to this health endowment needs to be studied. -those is notinh ty healthy or residing 5.51 Anoter soume of sanple selction that is in unhealthy envromenb are less lkely to sur- more subtle but potnty as important as morvive. Consider, hypoeticay, a health progm tality sdection is fertility seco Health prothat npedonately affects survival but not the grams (and food prkes and wages) not only afhealth status of those who survie (oral rehydra- fect wbhch Infnts are alive to be surveyed but tion?). Suchahealthprogramnmayactuallyreduce also the fertility decisions of households - the average health status of the surviving popu- wheth pregnancy occurs or not. As with morlation (and have a negative coefficient in a re- tality, changes in prgams and prices reslting duced-form for health) by permitting the sur- from struun adjstmnentay alter the fertility vival of some low endowment children who decsions of househols, and the additional chlwould otherwise die, while not appreciably aug- dren born, or whose birth is averted, as a resiut ment the health of those who would other- mayhavenonrandomhealthcharac eris. Pitt wise survive. Ornyffinfantmiortalitywasuncor- and Ronzwe4g (1989) examined the effect of related with health enowmeb would there be fertility selecton on the estimated parameters of no survival selection since the average endow- reduced-form health demands for infnts in Maments of survivors and nonsurvivors would be laysia and found fertility selection to be sigrifiequal. Only in this limiting case does finding cant and that ignoring it resulted in lazv misth a programhasnoeffectontheaveragehealth a of program effects on health. Like of survivors imply that the health of those who the case of mortlity selecdon, crrect for the would otherise swvive is not enhanced. Thus, effects of fertlity selection requires the estimaknowing only a program's influence on the aver- lion of a reduced-form having the same regresage health of surving children may be danger B so as the health reduced-form (10)(with the ISO

108 eweption of the gender of the chld as this is not affect the abilty of pubic health fadlities to known to the household at the tine the feriity provide servioes, or if t)ey are aware of spifidecisiom is made). The masue of fertity to be cant changes in fee schedules Data linitations used as depetden variable shoud be the num- can also nuke tabulation a useful alternative. For ber of prepandesover a refree period. In the example, if the community level data available is SDA household survey, anhrpvoy,mtric data are ondy a count of publc health facilites in each avaflable for surviving chldren up to age six, so region that provide a particular health service t'hat it would be most useful to have information without mard to the quality of smrice provided on all pregnancies that could have resulted In a and other attributes such as queuing time, a polild in tat age ange. icy shift that keeps the count of facilities unchanged but reduces service flows and quality Redud-fon dmnd for halth inputs will not be uncovered by regression analysis of the community level data. If infonration on 5.52 As noted above, the reduced-form deter- budgets, stg or other indicators of quality minants of all (contemporaneously determned) and serviceflows areavallable from oter sources, houseold choice variables have the same set of even anecdotal information, investigators would rwereso as the healt reduced-form. A iear do well to link this Inormaio with information formuation of reduced-form demands for health on changes n utiliztion deived from the houseinputs is given by equation (11). Most of health hold surveys. inputs typicay indluded in household surveys are meased as binay or polycto us van- stinating the wullngns to payfor medial cae ables which require the use of binary probi, binary logit and multinomlal logit esimation 5.54 As was noted above, it is unlikely that methods. The caveats assocated with estmating usw fees for medicd services ratim the market. and interpt the health outcome reduced- lheopprtuitycostoftraveltimehasbeenfound forms described above apply equally for health in almost all studies of the demand for medical input reduced-form The complete set of health care in developng courdries to be an important inputs msured in the SDA Integrated Survey and significant determinant of medical care defollow: mand. If the community or household question- Consulted health practitioner for chedcup naire provides information on the disunce or * Csulted health practitioner for illness travel time to differentypes of medial care pro- * Conlted health practitioner forprenatal cuare viders it is possible to impute a price to medical (last 2 wee6s) care. Using the observed or imputed wage of an * Cosulted health practitioner for postatal individual as the opportunity cost of time, the care (ast 2 weeks) price of medical care provided by a specified -Freqncy of postnatal care (12 months) type of provider is the fee charged by that pro- * Choice of health pactitioner for Ilhnss vider plus the product of travel time and the * Child vacinations by type wage. Wages can be imputed either by esimat- *Place of vaccination ing selectivity-aorrected wage regressions based * Duration of brof children upon the subsample of individuals who would * Participation in conmunity feeding program report mauket wages and hours, or more simply * Freqec of prenatal care in past pregan- in the case of rural areas, by averaging wage rates des (bygender) within ineach village. Within village * Source of prenatal care and gendergroups, individual vaiation in wages * Household food consumption (see Chapter is liely to be small For dhildren who are typi- Seven) cally accompanied by an adult to the health care 5.53 To finish the section a few conments are provider, the price of medical care can be taken in order. It is not necessary that econometric as the price of adult time for the gender which estimation be caffied out for au these health in- most often accompanies clr (women) plus puts. vestigors should single out a health any return to dhild time that can be Imputed. input for intensive study if simple tabulations of 555 Estimating the price elasticity of medical the data show that its utilization has importantly care is accomplished as described in section 1.6 caed in the aftermatof strural adustaent, (reduced form demand for health inputs) above. if they are aware of sigificant budgetarychanges The imputed prices for medic care should sim- 101

109 Ply be added to the list of regressors. IJeed, if nfomation on food cdnsumption at the individthesepricesarecalculated, teycnalsobeadded ual level is not avalable in the househd questo the list of regors for inlusion in the re- tionair Asfood Inake is lkelytobeaprpmary duced form demand equation for health. Their determinant of individual level health, estnregression coefficient has the intpretation of tion of the health production function is probthe effect of a change In the medical care pice on lematic. One solution adopted by Pitt and a measume of health outcome. Rosezweig (1985) is to estmate the health pro- 556 An econometric issue to conider in esti- duction funion using household data: the demating models of medical provider choce, is the pendent variable is proportion of household il appropriate estimation subsample. For example, and the dependent variables are household agthe household questionnaire provides informa- gregates of health inputs. This approach is valid tion on the health practitioner visited for Illness, if the individual health technologies are linear for prenatal care and for children's vacinations, and identical (except for intercept shifters) across The fonmer two questions provide information household member types (age/gender). beon provider choice conditimal on a prior en- spective of the data used, health production dogenous (household choice) event - being ill funcdonestimation requiresinentalvariable or pregant.'s No one seeks "prnatalr we if estimation methods (such as two-stage least one is not pregnant, or seeks treatment for ilness squares) since the health inputs are presumed to if one is not }ll. But It is not liely that the exis- be endogenous. The instrunental variables are tence of pregnancy or illnes- is unrelated to the thereduced-formdeterminantsofbealthlistedin observed and unobserved deterninants d5 pro- paragraphs 5.32 through 537. Identfication revider choioe. As a result, estimating a model of quires that the number of red ced-form detmihealth practitioner choice when ill using only the nants of health in that list which are not also subsample of those individuals who report included in the spedfied health production functhneselvesllwilllresultin-selectivitybla The tion exceed the number of endogenous health appropriate procedure Is to estinmate a poly- inputs. Two-stae least squares esimation is chotomous variable model having as dependent comurrcdy included in standard econometrics variables a set of indicators for choice of provider software packages. if ill (or pregnant) but also for the outcome not ill" (or"not pregnant-). Prindples of analysis and policy Impliaions The eca of hwath inputs Esthnating the health 559 The preceding discussion suggests that an prduion funcaion analysis of the relationship between government policy interventions, health and other household 557 The health outcome and health input re- behaviors has stiff data requirements. Health is duced-fonn demand equations descibed above multi-dimensional and difficult to measure - tell us how social and economic factors influence anthropometric measures, although expensive to helth behaviors. As the model presented makes collect, offer important benefits. The imrtn clear, the reduced-form is not the effect of health of unobservables in the determination of health behaviors (health inputs) themselves on health imply that longitudinal data will also aid in ewbut rather the combination of the effect of health nomic analysis. Panel data also allow the analyst behaviors and the deteminants of that behavior, to estimate the effect of canges in govement The reduced-form regression codeffcients thus re- programs without the bias that non-random proflect the preferecs, beliefs, perceptions and in- gram placment (as described above) can impart formation of the household, and not solely the Finally, the selective effcts of mortality and fereffect of health behavios However, the health tility can significantly affect the accuray of the production function, which provides the direct reduced-form determinants of health. CDrrCtbiological effect of health Inputs on health can be ing for these sources of selection requires data estimated and is of policy interest in assessing not only on the health status of the (suring) the potential efficacy and cost-effectiveness of sampled population but also on mortality and a health program interventions. complete pregnancy roster of all won. 558 Ideally, estimation of the health produc Using nutrient intake as an indicator of tion function requires information at the level of welfare or health has some serious drawbacks. the individual for all inputs that affect health. Nutrient intake itself cannot be considered an 102

110 argument in the utility functon or even a good DedMwO(98) ghidan datandshand Ad=a Irdiator of welfare lthe focus on nutrients de- 09)forsri anka rives fmn its importance as an input into the 5. &nrealgovematep r productiop of health and te comnmon absence of tity-seva devdeoing CPn*ic% Hkds and Kublsch (196 adequate data on health. Knowledge of how foundtsodalexpdtuewasthmetprotedamon nutrient Intake maps into health Is important for the five catere of pmm t expenditure defined, and policyanalysis. Evencompleteknowledgeofthe more so n Oe lowet Incme mins PFr a througb matrix of food price elasticities may not lead t examation of th evidents on the Impact of economic adaolusions about the effects of price interven- justuton nelhealth andnurlon see Beh 098). tions if price changes do not unambiguously re- 6. Dsoasion of ssue oundig the _meas m sult in a rise or fal in the intake of all important of health will be deerred to 5.2as thr ou& 531. nutrients. Furtherm, to the extent that house- 7. Below we wil alo inrdu a purly randomn onholds that have greater access to health services poen to health the imtae of which depends on how or live in less sanitary environments also face heath is measured. Fr example, tis random onet higher or lower food costs (e.g., mral versus ur- wm be bvy larger fe measue of hel is adiciolo banareas),cross-sectonal estimaes of food price m indicator of whethea an Individual is "aid on any effect that ignrme the health infrastructure may give dy tban If the meaiw Is the hightof an Indvidual. lead to mleading conclusions about the conse- Although individuals may have diffing to bequences of food price interventions. comeil (their i, whete or not one isick in a shot rfr In addition, studies of the determinants of enca perod depends more an purely random events tha nutrient intake are necessarily at the household- does one's beight, the rult of a lfetime of health behavis level in the SDA Integrated Survey. It is likely and endowtmml As wil be made dear below, this ndm that households do not allocate resources equally ompont to health differs fo thw aems le and). in hat among their membes and sgo he liwiinsof wbieallethreeae unknown to the researcher and ts canpolicy for the well-being of certain classes of indi- e to the mero term o tatially related relatbiie, viduals (women, children) are not idtntified. thie pand ters are nd randam events to the houehold Health data are at the Individual level so that. The discsn below refers to houshod frm acreducedform demnud equations kr health ci- tivites for dairy of prtation but appiies equaly to any rcty provide the effect on the health of the indi- rmueld production of gwds and sevim fr the arket vidual by type of chnges in polby interventions such an weaving rats, procesig ad ote sma cott (pices and programs). as we a tudng b sevic aid 5.62 Changes in individual health have impli- ther vmcespoducedbyhosbaldmembes cation for labor productivity, wages fertility and 9. paly, the sdcoling and age of the hed of the human capital investment. These are partdcu- hoehold and hds or her pous e treated as eogenous layrdifficut issues to measure,butignoringthem vales to be induded In the reducdfm demand equn nay lead to serious under-estimation of the re- diam Theageandshooingofdkh&nareoftennotcnsidmns to investing in health. ered exogenous as the reflect rent fety and human captal dedsion by the household which arenot independent Notes fiom the he Input decisions of households. Indeed, the Education Analysis Plan in hptr Six explbdtly coder 1. For an excellent and comprehensive review of this thewhooingof dlldrenasahoicevariableof thehousehold leahre e Bee rman and Dealar (198c). affected by structural adjustment 2. This reut pert to uncompensted price elat1-10. Explktsisl testsof theseparabilityhypothes ties for househods whose food expendits were 8eate In developing country agriculture have been peromed by than thoseof 75Mpercentfall households. Forhoseoldin Pittand Ros eg 966) and Bemin 9), boths the 90th perdlde, six of nine pdce ddltices had non- usngdata m Indonesia and both notable torect sepaidenta sgns aow the nine nutriens bly. The Pitt and Rosezwe4g pap is of paricdar ree- 3. On the importance of sbtuion eascitie for the vanboere shne it used Winfation on the health of olvaevautin of nutrition based food policy see Thmmer 0951), tor td test the sepaabily hypothess. The test was basd on Tbhmer, Falcon and Peason (1983), and Pitt (1983). Par an the flowing impition of separability: if perfdct market Aini asstudy,seestauss(198). emt for tne services proded by farm fmily otvatr, 4. The relatio p between productity, wages and te the heihissofany of Othn (which wasd wn health or nurtrent consumption has also been cmfimed by to significantly affect labor supply) should not affect farm Stss (196)usingdab for farm ho lin Siena lerneo, profit as market substutes are available Benjaminis sepaa 103

111 bly testwasbasidonarelatedtpcato houoldsize tatprematuren ateddfitytmoepnetohelt and comstion shud effed family lbw spy but not problems duwing her life Oth nfnt anted to fu term affectfanm profif labormakes are wellf i NeorMal Intuesve care s lely to have limpact on the 11. A simlar condusion about sf-rptd healtb Is healithofmostfullterm infants Aregressnhavinganmeabo draun by Wolfe and Behrman (196. ure of Wnt he as depedent vadable and the ao -ssiblt- 1P Farm or no-fm enterpbise profits shuld be re- Ityofintenslveneona_ta tenvecaefaltieasargessor psaced by ther -orm detmts of pfits-prices is lily to show that neonal Inesve care ities hawe a of nputs (Indudinglabor) and output, as well as quntities detrental effect on infant hedth. ne e re in of land capial and aoh fixed actors - f it is beved iat weftdent rdeeds the infhuce of neoat inenshre care on househ_ld production dedsions are not separable from con- tw surval of low endowmet infants. These low endowsumptn dedsions. ment infnts, bysrvivng reduce the average healthiness of 13 On e treatef this]prbl see Deaton (198). the sawmpled ( in popu It is dear that these 14 Howeve, this chae in the availability of hat filites can only mpove te heat of any inant coniserves may not be captued by all meawsr This point is tnal on its survival The negaive sign dsmply reflects the fur dcumssed bdow. sedive effects of mortality on induson into the regresson 15. lntaction is meant to indude not just the produts sample-surviving infaint of rerss with other regressors but also the product of a 17. SIme the dependent varile b binary in this aw regressor with ibelf (quadratic temms). (innt suvives, Ift dies, a probit ot logit regresion is 16 An example much more applcable to developed apprprats untries concms measing theeffect of intensiveneal Strldlysealungthecholeof heldsvsana care failites o some measure of the health of a samnple of provider Is ndition on there being a dild in the house (livin)infhnts. As welmown, a primary result of neona- hod - that is, prior fertility.-which shoud be nsidered tal intensive ce has been a damataly Iuesed rate of an endogenous oice of households as wel survival of very preauhe hants. ittoalsowe esablised lft

112 Analyzi'ng human resourc effects: Education mme plicy Issues members of poore househols,.lftuou not necesariy the poorest households in partcqlar, 6.0 EducAtin can occur in many forms. Usu- are thught to be most vulnerable to suc effects alfly emphasis Is placed on formal education 6.0 ecause of such complexfitie, In a sense through school because this is one of the most the impact of ad jutmenmi policy on education can Iptatform of education, becauseit is a miore not be considered in isclation The codncern of easily observed form of education thian i~~~~~~~~~~~~~~~~~~~0 are many othe Analysis Plan in this Volumte have inpowother and because It is a form of education that teknt implications for th~ determinants of schoolis fthught to be partiularly affected dlrvcty by Ing. For example, the determnnts of the povpolicy. For such reasons this Chapter focuse on erty profle; of emiployment and earnings, of education through formal schooling. But some health and nutrition and of the role of women all attention also will be given to othe forms of havelmprtantlomplctlnforeducation. 1 And, education - for example, through formal and perhaps with a lag, education will have imporon-the-job training and preschool and outside- tent implicaion for all of thes othe outcomes of-schol education in the household. of interesl Moreover, aft of thes social dimen Education is determined by both supply slon that nay be affected by adjustment also are and demiandfactors. Adjustment poliymay have affected by the overal macro developnmets and effects on education through both supply and structural shifts in economic activities that mkay demand. In some case these effects may be. be enedrdby adjwustent policy. Furthefary direct and perhaps obvious. In others they more, aft of these effects have an important time may be much more hinhdic and probably less dieso,with longer-run outcomes perhaps obvious sinoe they reflct the filwtern of adjust- diffein substantily from short-un feffets not ment policy through a myriad of differnt type only In magntude, but possibly in dirctionof institutions in the eomy-frsfas, 6.04 Nevr d, given fth apparent im~or PSIrmeta oraizton,private and public tance of education in afttining a number of schools an rann programs and housholds, society's goals, it is usefu to attempt to sort out These institutions, in turn. may be affected by the probable impact of adjusment policies on adjustnmtpolicythroughdirectpolices-such education. This Analysis Plan. attmpts to mdilas reduction in school subsidies~, increase in cale how this taskmightbeapproached. Theloci school users fees or increased rationing of school of the proximtate determintants of education are position - or much more Indirectly such as vario usmcro institutios such as households through the Impact of reduced fiscal deficits on farms, firms and school. -Themajoremphasis of household incoam and thu on the cosmpmtion this pape is on how to analyze the Impact of and time use choices of households. Geunerly adjustm%ent policy on educaition through such

113 prodmaute detminants, to which pargraphs 630 omic ments, hey may act to increase the through 659 are devoted. But befoe exmndng supply of such traiing even In th short rnl tids micro behavior, some consideration of the 6.06 The demand for education also is lilely to no-nesosettinginwhdchitoccurslsrelevant. be affected substaily by adjustmen poly. aragraphs 64 tbrough 6.29 discuss tlis seting. The major effect in the hort run usually is pre- And to motvate the whole Analysis Plan, it is suned to be a reduction in the demand for usefultoindlcate themainpolicypreoccupations sdhooing because of a short-n reducion in pertaining to adjustment policy and education income. To what extent this affects the demand Decausethesearedependentonthetimehorizon, for schooling, of course, depends upon whose it is usefu to distinguish between the short run income is reduced, with the strongest effects ocand the longer run for this purpose. cuning if ftose at the margin of schooling continuaton decisions e pein womereduction Policy inteknentions that are likely to influence There also may be reduced deands for schooling education in the short run that result from reduced health and nutrition of chidren, and thus less capacity for gainhg from 6.05 ThenostobvliousofthesepoUcyinterven- attending school or taining programs. There tions are ones that directly affect the supply of further nmy be reduced demand for education fomal schooling and baining programs. Adjust- because of time allocation decisions so that the ment policy usually involves short-run cuts in opportunitycostofchildrenestimespentinschool governmental expenditures in order to reduce increases relatively because of the increased iinfiscal deficits. The general cuts in governmental potnceof their time spent in either economic or expenditures may include cuts in subsidies for household activities For example, if women formal education and training, which in tumnmay spend more time in labor market activities to c-ause ncreased efficiency in the use of resources offset reductions in household income from refor such purposes, reduction in the quality of duced employment of men, older chfldren (parsuch education, reduction in the noneducational ticularly girls) may be called upon to fulfill - amenities assodated with education (e.g., school ous household activities intead of Soing to school luches), reduction of the quantity of such edu- But the same example also points to the possibilcation (perhaps accompanied by increased ra- ity of the opposite effect if women spend less doning of such education), and/or incrases in time in economic activities due to reduced emthe uses charges or prices for such education. ployment options or if children have less attrac Iceased efficiency of the use of resources used tive own firm/farm or labor market opn Also, in the production of such education may occur on the demand side as on the supply side, there without any diminution in the quantity and the may be a short-run increased effect if the adjustquality of such education. Such efficiency may ment poliincreases the expected private rate of be attained within a given schooling type, or by eturn to education due to expectd improve- -alocating resources across schoolng types. ments in the medium- and ionger-run prospects The ltter may occur at the same time that user of the eonomy if the adjustnent program is succharges are increased. All of the other changes cesful. The explicit adjustment polides that are likely to cause a reduction in the quantity of might have such effcts through the demand side educato thequalityof educationorboth. Who include any adjustment policies that may affect Is affected by such changes depends on the exact employment, wages and relative prces faced by nature of the policy changes. But in most soie- the household. This means, effectively, that virties the poor are politically less able to protect tualy aul adjustment poicies may have such efthemeves against such changes than are those fects, from devaluation to fiscal retraint to price who are better off. Note that while nostof these and wage policies. changes are Ukely to be negative, there are possibilities of changes that are positive on the supply Poliy inentiom that are likely to affect side due to adjustment policy. For example, if education in the longerun finns that have trainng prognrs with a long gestation perceiv that their medium- and long The longer-nm effects depend very much run prospects are improved by the adoption of onthe degreeof success of the adjsent policy. adjustment policies, either because of expected The more sucoessful it is, then as compared with sectorwal shifts or because of expected mcroecon- the situation that would have prevailed without 106

114 the adjustment program, probably the private two majorelemenbof themesosettflngtatserve reun to education of most types and the sup- as wonduts for fte inpact of the macro adjustples of resources to be used for education wil be mwnt polcy on the mco dt ton of edugreater, but the opportunity cost of tine spent in cation and then discuss the empc se of educatlon rather than in productive activities also linting economic adjustment policy to these conislhkelytobe reater. Geerally theformereffect duitsa is expected to dominate, though it is far from obvious that it wil in all cases. The relevant Mar demenb of themso seing th tnmsmithe policies in this case are likely to be the whole set impact of adjustme policy on the deterinion of of adjustment policies once again since what eduction counts are realzations (that condition future expectations) for the economy as a whole As noted above, the meso setting relevant for the detewmnation of education by households The macro-meso settig for the impact of andohrrelevantmicro institutionsindudestwo adjustment policy on education broad conduits for the transmission of adjustment policr to these miro institutions, markets 6.08 The micro level behavior that determines and infrasucture. These are now eonsidered in the supply and demand for education occurs tumr within a larger macro odntext that is affected by 6.11 Markets: Marketsfurthercanbedharacteradjustnent policy. The meso setting includes the ized in a number of ways. Probaly most useful conduit mehnism that transfer the effects of for the present case is to distinguish among facadjustment policy at the macro level to the proxi- tor, product and financial markets, with the recmate de of the micro household and ognition that each of these types of markets may other relevant micro institutional behavior that be more or less formal (in the sense of being directly determine education. On a general level subject to explicit contracts and govenmental the meso seting is likely to be affected by adjust- regulations). mont policies in two major respects First, by al Factor marets are important in at least tering the markets in which micro entities func- three respects First, for those many poor who tion SecoWn, by hanging the infrastrucure in work as employees, such markets are the major which they operate. The major adjsbnent poli- source of their income, which in turn is a major des that are likely to have effects on education determinant of at least some forms of education, through these meso variables are likely to in- including formal schooling (Chapter Four). If dude cuency devaluation, fiscal and monetary adjustment policies change wage rates or emtraint, and htermtional and domestic price ployment options in such markets, these nay libealization. A disusson of the fmnacts of lead to changes in the demand for education. If such policies on genral mnacro aggregates is not the initial stage of adjuste nt involves rernchin the scope of this paper.2 2 But It is useful to note ment, theseffects initially may be negative. If that such effes may be complex, depending on the adjustment program is successful, these efthe combination of expenditure switching fects are presumed eventually to be positive. (thrugh relative price changes), expenditure r- Second, changed wages also may have a price duction (through income effects), and changes in effect on thedemand for education in addition to expenditure composition that occur as part of the the income effect just desribed. One dimension adjustment policy itself or are induced by that of the price effect is the opportunity cost of time policy. And, once again, au of these effects have spent in education instead of in the labor force. If animportnttimedimensionsothat,forexample, adjustment polky results in an initial decline in the immediate aggregate effects may be labor market demand and in mal wage rates, the contractionay, but the longer-run ones expan- opportunity cost of time devoted to education sionary If the adstment program is successful. declines, though if the adjustment policy is suc The puposes of this section are to set out ceasful eventually it increases. A second dimenthe most important dimensions of the meso envi- sion of the price effect refers to the expected reronment in which therelevant entities for educa- tun to investment in education in termsof labor tion deterdnants funtdon and to review how market rewards. These presumably rflect exthese dimenions may be affected by adjustnent pectations about labor market developments afpoliy. We begin with a brief discuson of the ter the temination of the educafion, and thus 107

115 may rder to som tm in the f& f the all reduction in aggeate demand, though the adjustment poliy is perceived todhave some rea- price effect of devaluation wors in the opposite sonale probability of sucessuly Improving the dirction for tradables, some of whi (eg., basie econony overwhatftothorwlsewouldhavebeen saples) may be very importnt in the conswm without the adjustment progrm, then it would basket of the poor. Forth, a product price of diseem liely that epet rtuns on edution rect obvious relevance is the product price of investnents would inrease. However, this ef- education. lhe impact on this pieis not Immefeet may be damened by any initial negative diately obvious. For formal education, often adimpact on labor markets snce typically curaent justment policy involves a reduction in subsidies, expeiences enter into the forma of expect- sometimes the introduction of (or increases in) tons about future outcomes. Third, factr mar- user fees, and sometimes increased rationing of ket dhanges also nay affect the upply side of positions in public schools (the st two of which education. If initial restiive policies result in may be equivalent to price increases, broadlyan intial reduction inrealwes of more educated defined). On the other hand, at least In the short indhviduals, for example, there may be an initiad run, the possible reductions in the costs of the reduction in the cost of the most Important (at budgetaily most important input In the educaleast iinancial terms) input for supplying edu- tloa prcess (ie., more-educated perso) and caion. With a successful adustment policy, of of greater efficiencles and possibly of expanded course, over time this effect probably will be re- scholarship and loan programs may result in a versed by increasng demand for nmre-educated decline in the real market product price for edulabor. 4 cation Product marketshave at leastfourimpor- 614 Financial makets may have important eftant possible dimension for the concen of this fects through facilitating income production and study. First, in most developing countries a investment in education on the demand side and nunber of the relatively poor generate inome through facilitating the supply of education on primarily as smal-produces in agriculture or in the other side of the market as well Typically the infomal sector. The product prices that such adjustnent policy attempts to make such marproducers face are an importnt detemnant of kets more efficient by increasing to the opportutheir income- The ovea adjustment policy is nity cost levelsreal interest rates in formaedit likely to have a negative effect on the incomes mankets and by reducing rationing in those mar- Initially (once again prbably vsed in the kets. The rise in real interest rates, if anything longer run if adjustmentissuccessful) if tere isa would seem to discdurage investment in educarduction in overall aggrate demand and if tion (as well asotwerformsof investnt cets that reduction in other parts of the economy paribus. But geneally there has been very licauses laborers to shift into such self-employ- itedaccmesstosuchcapitamrketsbypoorpeoe, ment. But that is not the only effecl For certain prticularly for investments in education. So the important segments of this part of the economy effects depend in substantial part on the extent to - e.g., those involved In the production of which adjustament policy eforts in the fonmal tradables including basic agricutural staples - credit markets spill over into informal markets the aggregate price effect favoring such products and on the extent to which there is fungibility in due to devaluaton and trade liberalization is household resources (under the assumption that lwy to work in the opposite direction, and may even in the hifoma credit maet loans explicmore than offset any adjstent policy-induced itly for education are not common). Oce agan, decrease in aggregate demand even in the short the net effects are harely obvious. in the pirun. Snd,paaleltodiscusioninthe prious adjustment situation, for example, evwn if real paragraph, therealsoisapriceeffectforsuchself inerest rates were low (and credit rationed) in employed that affects both the opportunty cost the formal credit market, interest rates nuy have of their time to invest in education and their been quite high in the more revant infrmal expectations about the returns to education. 8 market Moreover, efforts to railonalize credit 7haS, a major ctor affecfting household real in- markets may cause an ongoing reduction in come of the poorand trefore their demand for interest rates in the infomal marets, and any education is the rtive prices that they face for short-nm restaint in aggregate demand may inconsmpton items Such pries are ly to dew duce a fall in interest rates as well die nitially benuse of the probable nital over Inf wcist isused here 108

116 to refer to publy-provided physical capital and on the provision of educational Idnrastctur puby-pvided servces. Inbroad temsinfra, Moreover, it appears at times that there are struture can be subdivided into social and eco- threshold effet with regrd to maintenanoe of nondc fst such capital stok with the esult that the aost of 6.16 The mwst important sodal Inrstructure delays may be considerable in the loger rune that affects the d mon of education obvi Though such educational Infstuue is ously is that wlich directly Is used in education, the mostobvious part of the socl infrastructure paticularly publicly-provided formal scooling to affect educational investments, it Is not necesand training. The availability and quaity of both sarily the only part of the sodcal nfucre the capital stock and the crrent inputs (most that is reant. Reductions in other sodal sewimpotantly the staff) ae critical deterdm ts of ices may have a negative impact on education, the amount of education and of Its quality. The either because their provision is coupled with the current inputs into the process can be altered provision of education or because they work fairly quickdy by budgetay stringenies, rest- though households to affect the demand for ing in inadequate avaiability of materia inputs education. Exmples of the former include food such as books and of persomel inputs such as and nutrition programs that are provided thugh teaces. Evenwithdvilserviceregulationspre- the educational systeu, such as usdized or cluding the rapid dismissal of teachers, for ex- free school meals. Reductions in such programs ample, nomal attrition without replacements can in themslves may reduce the attractiveness of reduce the numbers of teachers (often quite ca- sending children to school because of the lesspricously from the micro point of view of indi- ened tied (to schooling attendance) real incone vidual schools), and turnover with less quaified trnsfer in the form of food to the faiies of such replacements can reduce the qualty of such chidren. But reductions in nutrition and health teadhers. Reduced real wages and salaries for programsmayhaveanegativeimpactonschoolteachers and other staff as part of budgetary ingattendarceevenif suchprogamsarenot tied stringency may tend to Increase such attrition to school attendance, both by reducing the real and tunover. But such effects may not be so income of the former recipient households and large in the short run because ovall macro by ireasng to them the effective pries that strigency probably diminishes the Immediate they pay in money and/or time for helth and attrctivenessofalternativeemploymentforsuch nutrition inputs, which in itself may reduce the individuals. Even if such real wage dedlines do productivityof timethatchildrenspendinschool not cause an exodus of experied teahs in and thereby lessen the incentive for such chilthe short run, they may cause morale problems dren to even be in school.6 Reducdons in other orencouragemoonlight hatreducetheshort- socal ifrastrure, such as in more genera run efforts of tachers. They also may cause welfae provmisions, are likly to have a similar longer-nm exodus of epereed teacwrs and income effect on the demand for education, less success in att higher quality new ie- though not a particulaly inportant price effect cruits if the adjustment policy is reasonably suc Changes in economic infrastuctfue due cesaflflfmanaggregatewpptivesothatote to adjustment policies also may have indirect efopportuities for relaively-welleducated indi- fects on education. In the short rmn these we viduals improve, pariculaly if real wages for lily tobethree First,reductioninmldntenac teachers do not incease rapidly enough Though and n expansion of traportation and conmuthe Immediate impact of any budgetay strin- nication systems as part of efforts to reduce fiscal gencdes is lile;y to be mostly on current matel deficdts may incrase the time and monetary costs inputs and on staff, there maybe some Important of travel to educational institutions and thus efeffects even in the short rn on physical stock. fectively increase the price of such attendance to Sdhools and taining institutions simply may be actual and potential users. Second, the reducion closed, which may appear to be a sensible ratio- of such activities is likely to reduce the demand nalization from an overall point of view, but im- for laborer with negative impact on the income poses at least a tine and transportation cost on of houselds and the accompanying reduction those who would have utilized such kicihties had i their demand for education though the labor they not been cosed. Mainten, moreover, market (as dcussed above). Tid, the adjustoften is reduced in such, which ment policy may affect the expected longe-run may have a delayed, but sill substantial impact quatyof theeconomicinfrastructure and threby 109

117 alter expections about general ewnic development and thus te ewomic retun to educareqrs an eonomy-wide pespective and the capacty to contwl for the counterfactual sitution. It s hard to know a priori in whh dkrec- ation regarding what would have happened tion this effect is liy to work si reasonable without the adjusmt policy. If the county expectations nuy be for an Initfal deterioration in eriecd a omonodity boom or bust with r#- economic Inr*astructure due to short-run fiscal gard to its major primaiy commodity exports at stringencybutforeventual mprove ntsinsuch thesame time that itwasatemptto undetake inastructurerelativetowhatitwouldhavebeen the adjstment policy, for example, it would be without the adjustment policy if the adjstment misleading to look at the impact on the meso policy in fat succeeds in longer-run economic variables for the determination of education development-related goals. In the longer nm the without conuroling for changed world conmodadjusmt poliky can have similar indiect price ity markets. and Income effects on the demand for education, 6.21 Ideally one would lke an evaluation of depeng ccially on the degree of success of such effects for each of the major pso variabes the adstment policy. that are suggested in Frae for each of the nmajr types of households that are likely to be of Asing mpiric liks between macro adjushnent primaiy concern because their locatio in the inpolicy and the meso seting for the determination of come distribution means that their decisions reeduation garding educational investments are likely to be 6.19 The empirical assessment of the links bep ytculary at risk due to adjustment Hopefully some important elemts of the tps of meso tween macro adjustent policy and the meso data that would be desiable will be available settingforthedetermnationofeducationisdiffi- fom the other SDA studies of the country, parcult indeed. This is so for at least seven reasons. ticularly those on the poverty profile, and em- Fist, a number of the effects of adjustment policy ployment and earnings. on the meso settng descibed previously, work 6.22 To evaluate the impact of adjusment poltheir way through the complexities of the overall icy on the meso variables related to the determieconomy in order to affect factor, product and nation of education, ideally one would use an cedit markets. Second, some of the effects work economy-wide model of the relevant economy through altering expectations about returns to that inludes the explicit details of the adjusteducational investments, and such expecttions ment policy and controls for other shcs and usually are not known well by policy makers and that dewrmines endogously the meso varables alysts. hird, there is a time dinsion to the of interesl Such an evaluation can be fadlitated efects, with substantial differences and possibly enormusly if there exdsts usable and sensible evenmemasineffectsoverimeoufth, someof economy-wide modes that incorporate all or most the important changes in the meso setting maybe of the expicit policy changes that are part of the quality changes rather than quantity changes, adjushtentpolicypackage,otherprimarysowurs which usually are not observed vey well em- of macoeconomic s (e.g., changes in the pricaly. Ffh, some of the changes may occur intenational terms of trade), and endogenous thrugh rationing rather than price changes, and delermination of the major meso variables of the fonnermaybehardertoobserve or to control relevance. With such a tool, c n exin the estimates than the latter. Sixth, the whole periments could be conducted to evaluate the evaluation presumably should be in comparison impact of the adjustment pdicy on the major with what would have occurred without the ad- relevant meso variables. In the best of all cas jwstment policy, not in comparison with some such models may be available to detemine soe ideal orunsusalable situation. Sementh,in many of the variables of interest For example, ther cases previously exising analytical studies and are some models such as those in Dervis, de Melo tools, such as applicable economy-wide models, and Robinson (1982) and Adelman and Robinson are likey to be limited or noneistent. (1988) that generate some of the importnt vai- 620 Nevertheless, theissues areimportant and ables regarding income distrbution and factor therefore it is desirable to proceed as effectively and product markets. If such models are availas one can with analysis of the relations between able, they should be used t? esimate the Impact adjustment policy and the meso setng for the of adpusanent policy on as many as possle of determination of education. Such an evaluation the varables in Frame 6.1. But in the best of al 110

118 cases such models as are curmny available do Frame 61: Suggestive lit of meso vaiables not permit estimation of the impact of adjust- for SDA analysis plan f education ment policy on aul of these variabes Moreover, in most cases such miodels are not readily avail- 1 Mu able at all. Since the construction of such models 1abot or the adaptation of existing ones is likely to * wage ra ow-smied4 higher-smd as bas for require resoures far beyond those available for peed retutis to educaton; teachers, trs;by the study of the impact of adjustment policy on se, regi_o)nime) education, some alternative approac:tes gener Employen/Unemployment by ill ally are required to estimate the impact of adjust- N* a usos ment policy on the relevant meso variables. a_") 623 In the absence of such tools, two modes of * bt for education (pries and avaabties) analysis are recommended as components of a 12 Pdwts second-best approach. EAuatio 624 First, * Sdioollng price and availbilitts for the meso OInduding variables of interest It exg an icm s rino and otier would be desirable to undertake estimates of the e o,, deviations from the secular trends dunng the * Tnagpripes and avalailittes adjustment period. The relevant questions to ask Mor nsunmpton item are (1) are there significant deviations from the * 5Weandoh odpim underlying secular trends and (2) if so, do they NOnood prim (eg~ dotfg,f housgn health appear to be in a pattern that suggests that the Maor producs of small entprises-pzpies deviations in the relevant meso variables may 13 Fimd,J have been caused by the Jeviations in the adjust- Fornal: niminerate4terns, avaitabilitles ment policies instead of by otier macro shocks? r s availabilities Simple regnssion of these variables (or of their 2 u logarithms) on time trends and on dichotomous 21 $Ochl variables for the sructural adjustment period Educaion provide a means of evaluating whether there have * Sdlng: quatty and qualty (eg, aers/ been significant deviations from the underlying eudmi, eduatio of teacders books aid ohe trends during the period of the adjustment ef- * Ti; quateand peqruudaity fort Such a procedure permits the separation Noquauca nqity mad quality of hean and from ongoing trends of events that at least are weiae servics coteminus with the adjustment policy. There- 22 Eoxwwk-awst adndbt-eurnoante uum=d fore it lessens the posibility of misinterpret ig for example, any ongoing deterioration in meso variables due to the previous situation that called Empsynlentked for the adjustment effort as being due to the ad- Exteon (aicuua othe) justnent polcy. It also lessens the possibility of mis-attributing an impacto the adjustnt policy If there is a one-year change in the reported value fachtal that (at least eventually) is too optimistic of some relevwtt meso vaiable from the period about what the situation would have been oter- Immediately before the adoption of the adjust- wiseandthustoopessimisticaboutthelmpactof ment policy to the first period of the adjustment theadjustmnprogam While the bias towards policy by incorporating information about a excessive optimdsm (at least eventually) about long time period. 9 the cou aal situaton seems liy if in Whilesuchaprcedue Islikelytobemuch deed there is a need for the adjustment policy, it more systematic and satisfactory than frequent does not folow that there is such a bias for the efforts that merey selectively eyebal the avail- initial part of the adjusmnent period in such a able data, such a prooedure has its limitatons: counterfactuat. If emplomt was high prior to (I) The presumed counterfactual is continuation the adjustment policy because of unsustainable of the secular trwds, which in some sense could govmentl def1cits for example, then eventunot have been maintained for long since the pre- ally - but not neesarily at the start of the vious situation was not sustainable; for this rea- adjustment policy-such employment probably son, if anything, such a procedure uses a counter- could not have been sustained. At most such a ltl

119 procedure indkates modaton not causality, and Mebowmicro analysis of the Impact of even understanding the association may be adjustmet polihcy n educaton somewhat difflcult if theme are lags.in the impact of adjustment policy. 628 Theprevioussectiondsussedthemarket 626 But despite suich limtations, in my con- and infrastructure dimension of the meso settexts looking at the deviafions from the secular ting in which the deteminants of education octrends of the relevant meso variables nay be a cur and how one might estinate the impact of cost-effective way to evaluate the impact of ad- adustment policy on the key variables in the justment policy on such neso variables and to meso settig Paagraphs 6.01 thrugh 6.27 sugavoid errors that easily can occur from selective gest a number of hypotheses about the Inpact of eyeballing of the data or from proceeding on the adjustment policy on education, as transmitted basisofapririsuppositionsm Theresultsof suc though the maket and ifastructue conduits an investigation indicate wheter the changes in of the meso setting the meso variables were significant during the Hypothwis 1: The income effect reflectng develadjustmentperiod. If theyweresignflcant,such opmet in both factor and product markets, is estimates also give an estimated magnitude of liely to lead to a short-run dedine in the dethe assodated change that could be used with the mand for education cetris parlbus for housenmco estimates discussed in paragraphs 6.28 holds whose inoomes dedie due to budgetary through6.59toestimatetheimpactoneducation. s e or dependence on production of 627 Seond,forsmeofthemesovariablesthere nontradables (e.g, many urban labor and informibht a priori seem to be parcarly strong links mal sector households), but even in the short run with particular elements of the adjustment poli- there may be the opposite effect for households des. For example, reducdons in real central that deperd on tradable production for their ingovrnental educational expenditures plausi- come (eg, smal1holders and rual labor housebly might have a direct impact on the avaiable holds that depend on production of tradable indicators of educational infrastructure such as staples for thir incme). In the longer run, if the those that are suggested in Frame 6.1. But at the adjustment policy is sc f, the generaly insame time a x% reduction in govenmental ex- ceased income is likey to lead to increased dependitures is not likely to have a x% impact on mand for educaton for most households, teris the indicators of educational infas be- paribu& cause some htemal efficiencie my be imple- HyqaUz. 2: The direct educational pdie effect mented, because rlative prices may change, and rlated to the cost side is lkely to be negative for because there may be substitution among inputs many households, but ambiguous for other in the "producton" of education. Moreover, the households. In the short run the time and moneffects of the reduction in central govenmental etay costs of education awe Iley to increase educational exendies pbably affect differ- generally due to reduced educatonal and transentiauy different types of houseds of interest portation infrastructure, increased formal due to their differentialocations and use of dif- schoolingandtrarguserhdrges,andinceased ferent omponents of the educational infrastre rationing of positions in public schols and ture. Therefore it may be usefud to estimate re- trang pgam, all flecfin bgetary strinduced-forn relations with the available india- gency, though some of these effects may be lesstws of educatonal infastuctureas the depend- ened with more time if the adjustment polcy is ent variables and with central governmental ex- succesful. Households and individuals that face penditures on education and other important in- incraas demands for thir labor in the shot run dicators of adjustent policy as the right-side (e.&, smallholders and rural labor households variables. Such rlations may provide a usefud that depend on tadable producion of saples for basis with which to explore the possible impa:t thwir income once again) and In the longe run of actual and of altemativepattems of changsof (probably most hsholds If the adjstment central governental educational expenditures policyissuccessful)havethenegativeprice/cost and other policies on education. effect reinforced through the higher opportunity cost of their me in tem of eonomic activities. On the other hand, in the shot run a number of poor households are kely to face reduced opportunitycosteof timedevoted toeducationbecause 112

120 of fte lessened overall aggrate demand, rein- ponents), the household also plys a mapjr role forced by shift in the composition of demnnd as te suppler of some foms of educato Pardue to rlative price dhft or ose involved in ents, relatives and older siblngs, for example, producoofnontrdabes Theopportunt cost help toeducateyou dcdidren. And probably of time dim of educational prices/costs, there is some imporant education that adults thus, Is likely to bein the opposite dhicon from recelve fom the idldren as well, though there the income effect. may be differential pereptio acmss the gen. Hypothsis 3: The pdre effect directly rdated to eratmonsabout trelativeimportanceof theflows the expected rate of return on educatonal invest- in the two directions. If the household also Is the mentsislikelytobegenea_ ypositivefftheeare locus of a family fa/flm, there probably Is expectations that the adustment polcy will be furthe education that accompanies the employsuccessful in Improving at least somewhat over- ment of household _mms in that interprise. all econoc prospects. The eeptos to this Sometimes, partulady in agriadture, such edugenwl statement would seem imited to some catkonaboutthelocalicro env_lrnmtappeas spealied form of ttinng for occpations in to be quite important in detemidng icrao producton activities in sectors that are lkely to household enterprise productivity. dedine because the expected negative price ef Before tuuning to the conceptualization of fects due to the adjustment policy more thanout- household behavior, two genr points merit weigh the expeced positive income effects if ad- enphasis. First, the definition of what constitutes juspten policy is somewhat successf in the a household is not without conowversy. Is a longerrun. Casualobservationsuggeststhatsuch household a coresdent group? Those who share spcialized training for such dednng sectors is a hearth? Those who shre other functons and not ikely to be vely widespread. resources? But for practcl purposes, the defini Tle ptresent secton is the heart of this tion of households that mates to the resarcher Chapter how to analyze the impact of adjust- isthatusedinthesveybasisathand. SWond,the mentpolicy,throt hthelnducedchangesonthe household is not the only institution in which meso setting on de iacro determination of Individuals are involved, nor the only institution schooling and how to test r ting hypothses. thatmayaffecteducationaldemand. Individuals At the micro level, as emphasized in paragraphs are involved in a multiplicity of networks and 6.01 through 6.07, there are both demand and relations, including (but hardly imited to) relasu1ppl sides of the deiaton of education. tives, friends, religous associations, work rela- But to the extent that much of the supply side is ions, tribes, clans. AU of these may affect the determined by governme poly (as for for- det naon of education. But generally the mal public sdcooling and taiig), Important household is the most Important institution of and perhaps predominant comp ts of the revan for this purpose, and to a certain extent mpplydetemnantsaheadyhavebendbscussed the roles of these otw hestutions can be capwith reged to the mes setig They basically tured by resource trans to and from the affect the Infrastructure faced by th household. houselold. In fadt the avaflable data do not Therefore this secion focuses on the demand for permit much presentation of these other instieducation, tough such demand Is cditona tutions beyond the resource tansferol on supply c tis and there is some dis- 632 he standard economi c uon cusson of supply coderations. Sincethelocus of houseoldbehaviorindludesa setof allocation of demand de;sions concering education is rules, househod production functions, budget usually cenered Ln indhviduals who share a, in assets of the housenumber of resources withdn a houshd, the hold, and ex nos pres and infrasructure. household is the focd uit of analysis. An example of an allocation nre would be td maxidmze the satsfactin of household prefer- The houss rok in the d ination of ences or to enae in bargining among houseelucation holdnmenberstodetermineallocatlonof resurces and of tme uses among dm. An example of a 6.30 As noted, the househid plays a major household production funeton would be that role in thedeterminationof thedemand foredu- the helth of a cild is "produced" by the food, cation of its nmbembr With abroad defintion of water and health care that the child receives, the education (not Just one limited to its fomal comr- ds genetc endowmnts, the household and 113

121 community, the health-care skil 634 The asic mb rch stwry has the followthat the cidws mother has (reflectg both her ing four steps formal education and the haits and knowledge I. Prepare data from the SDA household survey she has acquired over time) and the time that she and from other sorc that permit estmatin of devotes to health care of the child. The budget a set of rlations of the gal form of equation consaint is a 'full inc(m constaint that indi- (1) for all of a number of outcomes related to cates that the total available household resources education that are observed in the snvey. (including the tme of household nm _bers, as 2. Esimate relation (1) from these data for all of wel as physical and financial resources and the relevant outcomes. transfers to the household) must equal or exceed 3. EsImate the changes in the right-side mew the total use by the household of such resources price and infastructre vaiables that are due to (purchases at market prices, me uses including the adjustment policy for a given duration of tat those relabted to education and to econoic ac- polcy by relating the chawges in the variables tivities and househol production and leisure, used to esfimate the alterative versons of equaand tansfers from the household). The predeter- tlm (1) to the estimates of the anges hi the mined asses of the household include physical meso settng induced by the adjustment policy and finandal assets and human resource assets, that are discussed in the previous secion. which in turn relect previous fonmal education 4. Use the following diffeced form of the and training and usually unobseved endow- estimated relations of the form of equation (1) ments due to factors such as genetics and child- and the estimated changes In the right-side meso hood environent of current adults. variables from step 3 to estimate the impact of 6.33 Application of the relevant resource allo- adjt policy of a given duation on various cation rule leads to reduced-fom demand rela- outontes related to the detminatlon of educations for each of the variables determined by the tion household, includg the time uses of all household membes and inluding among those time A Z f(a P, A), (2) uses the me spent investf in education. These left4sdevaiablesmaybecontinuous,suchasthe where A is the difer operator for the reletie spent in education or test scores. They also vant duration of the adjstment policy under may be lhidted dependent variables, such as ad- consderation. mission to a given schoolin level or to a given type of sdhooling at that level (e.g., admission or 635 Most of the rest of this section is devoted not to public and to private secondary schoo}s). to elaboraftng in some detail on these four steps. The right-side vaiables i these relatios are the But before tundng to those detalb, it is usedul to variables that are given (or p nd or ex- emphasize five important features of elations (1) ogenous) from the point of view of the house- and (2). hold: assets, prices and infrastruchtie u 6.36 First, both relations refer to an outcome for a specflc individual, but both include on the 7=fi=(A, P, I) (1) right sidend actsisof all i- dividuals in the househotl That is, for the dewhere ZiJ ib the jth outcometernined by the temdntion of the time invested in the education household affectig the ith individual in the of small children or the probabiity of admission household (most imporat examples of which to different sdhooling levds, the schooling and for this chapter are the me and other resource other huan nesource endowments of al houseuses of various individuals in educational invest- hold members are included. This is the case ments, examination and adis- because if any of these predeterined human SioS to various types of schools); Ai a vectdor resources were differnt, the difference would of all of the predetermne assets of the house cause a change in total household resources and hold for the period of relevance (physical, finan- a re-allocation of household resoures that in cdal, hunun resources, gentic and other endow- genera mght affect the time invesed in educaments); P is a vector of al the relevant product tionof smaldchildtenortheprobabr ty ofadmisand factor maket prices that the household faces; sion to different schooling lvel toger with and I is a vector of all of the relevant inlfasruc- all otheroutcomesdeterminedby thehousehold. tuwe that the household faces. This means that for the exploration of the impact 114

122 of adjustment poli over a given duation on a identical incease in fily hime number of difent Idiensions of educational 639 Fogrth, any given estmation b for a given investms by a hoshold (e.g, education of time period. As the time period ln sthe small dhild admidon to leves of education chane in th right-side vaables in equation (2) for wlidch there Is rating, apprendtesps of al diffr. The time patten of such anges olderd traftofyoangadlts),the t- prsmably depends critically on the time patside variables should be the same even though tern and the success of the adjustment polic. If the outcomes dfffer. budgetwy stringency causes deter_ioon in in- 637 Seond, for similar reasons the pries n- frastructure and rduced real wages lnitaly, but cludealrxelevantproductandfactormarketprices adstmnt policy leads eventualy to betr Inand the infrastructure variables indlude all rele- frasture and higher real wages than would vant socal and economic r That is, have pvailed without the adjustment polcy, the prkes Incdude not ony the direct prices of then the short- and the longer-run npact of adeducation, but all product and hfctor prices af- justment poliy must be cowidered seprately. festing the household. These nclude the prices 6.40 Fif, the exdluson of any of the right-sde of all important consumptim products and serv- variables at the estinmtion stage may cause omitices items faced by the househld, induding the ted varible blses if such variables ae correprices of the opportunty costs of time of house- lated with obsved (and inluded) right-side hold mmers. If the household has a family variables Important examples of difficult to obfamn or enterprise, moreover, au the relevant serve right-side variables include abilities and product and factor prices for that activity also motivationsof differthousehold memersand should be indluded. u Likewise, the inftastruc- ratoning mles, both of whkh may cause biases, tureshouldindludenotorlythesodalinfsr for example, in the estimated coefficients of ture directly related to education, but also the household assets or in*ome I The suggestion social inre lating to health and other below to indlude among the outcomes admission outcomes and the economic infrastructure. examination peformance and admission success anges in the health infrastructure due to ad- Is nivated In partby the desire to deal directly jusftent policy, for example, may affect house- with rationing though this exploation will be hold resources and resource allocation (indud- freeofbiasonlyifsuchexaminationsametheonly lngthetimeforinvestmentineducation) wether basis for ratoning. If there are other bases for or not they affect the health of the individual of rationltinpariularcountrles,twouldbevery conern regarding a parficular educational in- useful for the researhes to collect information vestmet on the ationing procedures so that there can be 638 Thid, note that income is not included control for such rationing In the estimation."6 explidtly in either relation. However, income is inluded inplicidy because income is the return Step 1: Variables to be sd Xi tx mikro anal of on assets (imporly including time) and both h hos" demd for education the assets and their returns (the latter in the prices) are incuded. This treatment avoids the estima- 6A1 We now tun to details cncri the tin and interpretation problems that would cc- variables that are represene at a high level of cur if an income measure were included that abstraction in equations (1) and (2). Of course reected the eogeous choies of the house- exact representations of the concepts underlying hold, such as the number of hours worked in many of the variables are not available in many economic activities. It also abows, in so far at the cases. However, often one or more indicators of data permit, the representation of the posibihty the desired variable are available. If there are thatthelncomeeffectsvarydepelngonwhoin several such Indiators for a parcular variable, the household receives such income. For ex- it generaly will be desirable to explore the emample, some conjecture that the income impact pirca implications of the altemative indicators will begeateronchdldrnseducationif mothers since each may be mnesuring Imperfectly the receive such income than if fathers receive such phenomenon of interest. In some cs good income. If so, an Inrement in mother's educa- indicators for the relevant variables exist in the tion cewis paribus might be assocated with a SDAhouseholdsurveydata. Inothercasessome greater impact on children's schoolig than an manipulation of the SDA househld survey wdi increment in fathes education that resuts in an be reuired. In yet other cases - parar 115

123 with egrd to data on test scores or other vai- redy as would be derable in an ideal world. ables used to ration public school plaes - it is Because ratong of shool placs appreny is strongly recor d that the SDA household iprtat In many onmtts, especially for postsurvey be supplemented by comunity surveys pdnr y schooling, it would be desrble to supand otherapproaches to obtain the cessary in- plement the curent prototype SDA household fo n. I osiderifirst the dependent variables srvey with Winnatlon related to the outms and then turm to the right-side variables. Bore of swch ratiordn For exanmep If rationing s by consideringthe details of coern aboutindivid, scoes on exami ftons, I would be useful to ual variables, however, t is imortat to empha- know if each child sat for such examination in siw hateachofthe reevatvariablesfirstshould the past year and if they did, what scoes they be examined to see if the disrbtions are plau- obed (or at least whether they were successsible since errors easily can creep into statistical ful in obtaining admission). "" "I Because analysisquestonsshouldbeasked,forexample, sdhooling attendance vaies enomouly aaoss about whether the ranges of variables asen- diffet childrn and with differing cditions sible. U som seven-year old children are re- reatedtodhildhealthandparticipationinlhuneported to have completed five grades of sdhool- geerating activities, it also would be useful to ing, for instae, there probably are sone prob- collect infnoation on days absent from sebool lems in the underlying data that need to be xctl- for children who were enrolled in scoo in the fied before proceding, past year. 6.4 Since schooling attendanc and examina UIACTma ow cp r sew [cancal r - tion sitting and perfonnance all are highly age MENS AM or,i RELMAIn OUfCtI Recent edu- ependent and perhaps sexdependent, the first catioal investments and oth related outcomes step in the analysis should be to summize the (e.g.,adm sontopublc schoolatdifferentscdhol sample data controlling for age and sex. Fame levels) are of particular inter because they re- 62 provides a usefid framework for an initial flect recent deveopmets and because the vari- smnary of a nmber of televant variables by ablesi that will be available to represent the right- age and sex. Age is indicated by year in this side detenninantsof education for the estimation table, not some mre aggrgated grouping, since of equation (1). The education section of the there often are significant dfferencs between prototpesdahouseholdquesionnlae ntains two adjacent years of age regading schooling inormation about recent formal sdhoolng and attendance and exuaination peror,ma, parsomte information about apprentieships, though ticuly forages that are the tr ions between Itdoesnotinclude muchinomationaboutother one schooln level and the next schooling level forms of education. for muny shdents. Examination of these data 6.43 For household mebers in the 7-30 age wirl provide the basis for making sev deci range, it Is asked whether they attended shold In slons about subsequent analysis First, this exthe past 12 month. This is likely to be the single ainaton wfll be acrtical input into the decision most useful variable in the curent SDA proto- on which ages to concen=te. The ages of most type household questionare for the _ t Intrest for the present Analysis Plan are the ages of the impact of adjustment policy on education at whkh thee is subta varance of scholthrough the estimation of relation (1) and the use ing attendance and at which decisons ae made of relation (2). The advardage of this variable about continuation of schooling to subsequent over others is that it refes to the recent exper- levels, perhaps in part through some ratoning ence on which the other household and comm- mechanism. such as admission examrdnation per- Tdty sur data focus. However, since wheer fm If due to cmpusory shooling aftenor not a partcular individual attended sdhool in dance regulations almost everyone attends the the past year is likely to be heavily dependent on inial gades of sdwolng for example, then the whether that individual attended school the pre- agesassodated with those grades are not of nuch viousyear,inasensethsvariablestillreflectsthe iterest for the proposed analysis. If alrot no Impactof moledistantpastconditions. Withthe onecontinues into tertary schooling, on the other data obtained in the prototype SDA household hand, the ages assodated with such schooln survey, it is possible to explore some dhmsion also ame not of much interest for the present purof the possle impact of pevious onditions in- pose. What is of interest are the ages for wlih directly, though not to cntrol for them as di- there is substantial variation in scholing atten- 116

124 Pame 62: Surawy of Khoolla urclet - k mli SDA survey: rane, median, mean, standa deviation and number of observatons for fllowing viables byage and sexw GeRom dm. uim< ' * Everattendedsdtaol ut * H1ghesatgade * H4_8atpde. * 12 hest diploma * bntsfupeiqn? &ltfuaiih' * dglitgrief * SamVed/cdiegeaftendance Repeat 8ade? * Gradeaoldh * Nuberof 1peat * lmeoingtoand sndah * PiblWpra EduAubual p 4psd 12 daibs) * Cetiftet? * S&osndmeghatwtiefe S MIury * Cm~topauUo.stodparenasdAos * Evst_Wt type * Unif55fand5pcw%do&e * H*,estgrade * Boo and sodssupes * Rpetade? *Tanspotatimn to swl* Numberofreab * Po4d,boardandlodngg * Hgedst srtficto * Oeralse(dubs,extra ae * PM*i/ptMC * Otherin4dexpnes Tec1kai * Wbopaid * Erattnded * ScoldashIp? * }8 h Bestgrade * Vaeofsdoladdp * H_Ighst rlfiate Cwaimtelae of main o0apatlonto past 12 months T; y * Talg 'presea job? * EWratended * Trai durat * Yeas * Hoarsperweeklntribn *istttit * WhopdfortraiAng * Hgestdoina * Say werdwing tring? - * Howum slerylower * Ever been? *Howong Characteristicsofsnma y job In past 12 mdoh * Main dbde TraIni? * PeM? * Traring durati * Pemomwndboerd? -Thshdnainsa Mgb puieqtyaddlina vales * Evetal ended Ex_mkwdmpaD &tlleamd of_hnt_adm * Monthsattended (fwsd:nwmwdfbrs&anen0padba"yf0rmainsubject ps yea, but also for eni eyewif posibbe * Whethleinsdloo1oneyenrea_Dleand,ifso,In what grds * If In sehool in past yar, numberof daybsaet fron scho - Basedon May 9vwuLonSDAdrafths duas IaI. d Iot, uv.dtvam erswthiudvdyftwcamgous b. at eklttuff rps ipen aubawhlchnhohterf gs : hls knot i SDAbo ldsars4wbutltw=dbeva% toaddlt ff pubicscodlpsiss etl bymmemeowth an sim _ pmuieft it bed MbstJeedhctmutaio ouu udolngd Imsm Ias each <dlwiud Itd fred wilh rgad td t meudsm d. If esamlaton sersame got avaiable inskmbad aboutpisg kulfflingadl,ientms wllbe uuuful. ohw danceandforbwichtereasubstantialchages aitendanc, in Ighet grade wmpleed, in total in schooling attendance. WOnd, issmnyo expensehs, in a Olarohips, in who paid for sdw ftedatawilgivelnslghtlntowhetherothervar- hng expses, im shooling h bteupo in geables are of Interest or not for differin age-e pean grades, W days bsent fm school, in group For example, scholashis may be of sitti for admlsdo exadnatio,w in ccess oi teevae ony for certin a rge or schol admission examinations, and in cognitive level Third, these data will give an irdiation of achi In ligh of the often hypotsized whetr there are sex differenals in schooln ihgh etums to female schoog and the frequent 117

125 apparet difnc In the educational options fined relative to that for other individuals of the for males ves females (e.g., Schultz 1988b), same age and possibly sex by subtact ithe asking systematicady whether there are such dif- nean experience for the relevant reference grup fetences is useful. Careful examination of the defined by age and sex.'2 Such a procedure has version of Frame 62 for males versus that for theadvantage of controlling for age and possibly females should be enlightening in is regar. sex inaman er that makes It possible to combi Quite possibly such differences are small, if exis- observations across individuals in the multivaritent, in certain age ranges (probably the younger ate analysis discussed in step 2 below without ones), but may be larger and favoring males in further concem in the estimation for age and sex other age ranges. Therefore such examination differenes, but at the cost of assuming that once should proceed with sensitivity to possible there is this control there are no other age-sex changes in sex differentials with age. Beyond differences in the coeffcients of relation (1). visual examination of these data it will be useful Therefore subtraction of the reference mean for to test for statistical diffbee sbetweenthe sexes the sane age and possibly sex seens to be the for different variables so that It is dear where the prefrred way of modifying the dependent varidifferences are signfict in a statistical sense. 0 able to control for age and possbly for sex. Two, Fourth, these data will provide useful background alternatively, some or al of the estimated coeffifor the multivariate analysis In step 2 in other cientscanbe allowed to differacrossageand sex respects. For example, as noted above, there is a with the untransformed variables used as the question about how dependent the schooling at- dependent variables This has the advantage of tendance decision for the past year is on whether allowing for greater fiexbiity in accounting for or not a particular individual attended sdhool in differential effects in the coeffcients of relation the year before last. The infomation sunma- (1) by age and sex, but at the cost of estimatin& r.zed in this table regarding attendance propor- presenting, and interpreftng more relations. This tions by age and schooling inteuptions should cost, of course, is less the more that the initial help to understand whether or not for given age examination of Frame 6.2 leads to a limited numranges almost eveiyone attends school (so for ber of ages of interest The second altemative is such an age range there is not a problem regard- likly to be preferable unless the difficulties of ing past experience) and whether schooling in- estimation are very large (so that there is a great terruptions are common (in which case whether advantage to undertaking fewer estimates) or the or not an individual attended school the year sample is very small (so that there is a great before last is less relevant in the decision about advantage to combining observations across age attendance in the past year). Another important and sex groups). example pertains to the question of who paid for 6A6 The first panel in Fvune 63 (overeaf) gives schooling expenses; in particular, to what extent the dependent variables of interest, which are did individuals beyond the parents pay, since available from the SDA Integrated Survey (with that may influence the specification of the right- suggested additions regarding examination perside variables in the estimated version of relation formance and days absent from school in the past (1) for some age ranges. year) and with an indication of probable priority. 6A5 After such examination of the data, the The first-priority variables refer to the education dependent variablesshouldbedefined. In such a experience of the past 12 months, for which pedefinition a basic question remains how to treat riod there is considerable information for the age and sex differences. As is noted in the first right-side variables. The second and third priortwo points in the previous paragraph, considera- ity variables are of interest, but are measured for tion of Frame 6.2 in most cases should narrow any time in the past (not just the last year) so the considerably the relevant age ranges for ide- indicators for the right-side variables have much pendent variables of interest The examination of more masurement error given that the nightsex diffe mnces in poit three also may limit the side variables change over time. sex differences of relevance. Nevertheless, in 6.47 The second panel in Frame 63 gives smikmost cases there wil remain a number of ages of lar hnformation for the right-side variables that interest and some evidence of sex differences in represent the predetermined household assets, schooling attendance and related variables At the time and money prices, and the infrastucleast two strategies are possible for dealing with ture. The major general point about these vanthese. One, the dependent variable can be de- ables is that the last two groups depend much 118

126 move on communidy than on household chac- cedures should be used first for estian These teristlcs, and thus are more likely to be attned have the advantage of being relatively easy and fom mmunity quesionnda, though by av- quick to undertake and of bengt relatively robut eraging across household variables in a given to problems in the underlying assumptions In commwunlyonealsocanobtain Ifomation about the present contex, moreov, thee shuld not the nature of the communty markets and infra- be problems due to simtaneity and less-thanstructure. Theindicatedpriorltiesareordymnt the-usual problems with ondted variables bito be roughly suggestive about what variables ases However, some of the proposed dependmdghtend to be most important for the purpose ent variables are limited rather than continuous, of this analysis plan, though of course there may in which cases the assumtdons underlying the be substantial variations across countries, regions OIS estmates are not satisfied and OLS estiwithin countries and household types. mates may be misleading. Important examples 6A8 The following variable-specific point of such variables include the dichotomous ones should be noted. Por tlhe prederm adult for whether one was in school in the past year, educatein in fte houshld, education for par- whether one has had traning, whether one sat ents is indicated as first priority. But, as noted for admissions examinations for particular above, others may play a major role in the deter- schooling levels last year, and whether one was mination of children's schooling. ff the examina- admitted to particular schooling levels last year. tion of Frame 6.2 leads to the conclusion that For the relations for which such variables are the otes within the household often are important, dependent variables it would be desiable also to it would be desirable to treat the schooling of use some limited dependent variables esimators oer adults in the household as first priority such as probit, logit, or Tobit to see how robust varables rather than as a second pioity vari- the estinmted effects are to the OS violation in ables as is indicated in Frame 63. the undelying distrbutional assumptions. This is particularly likely to be imortant for variables Step 2. Estmation of relaions for the demntin for which the eans are far from 0.5, such as the Of the educatio indicato proportions of children at higher ages who attended school in the past year. For some possi- 649 Once the variables in Frame 63 are ready, bilities the alternatives are more than two - for the relations of the form of equation (1) can be example, whether one was admitted to a publiuc estmated with the altemative indicators of edu- school, a private school, or not adndtted. For cation that are given in Panel A and with the thesecasesmultinor alestimatesshouldbemade right-idevariablesthataregiveninpanelb. The with all of the possibilities considered in one natural starting point would seem tobe estinates system of estimates. of the form of relation (1) in which all of the right- 6.52, hfora given dependent variable there side variables enter in additively. Several points is the question of how aggregate the estimates about this estimation merit emphas. should be with regard to age, sex and household 6.50 First, al of the estimates shud be multi- type. This issue was discussed at some length variate estimates in order to control for the other above with regard to age and sex, and it was observed deterninants when obtaining the esti- concluded that It probably is prefeaame to undermated impact of each individual determnnant In take separate estmtes by age and sex in order to fact it is only to control for such possibilites that explore whether the coefficients differ across age the A matrix should be included in equation (1), and/or sex. Also there is discusion above about since in the differenced version in equation (2) it theposslbilltythatdifferenttypehouseholds(eg., disappears. That is, it is only because some vari- small}klders versus rural labor households verables in Aare likely to be correlated with those in sus urban labor households versus small-scale P and I that is necessary to iriude the A vari- householdenterprises) mayacedifferentdemand ables in the esimation of equation (1). Parental conditions beyond the ways indicated by the obschooling, for example, is lkely to be associated served right-ide vanables. For the 01 estiwith the quality of local schooling li I; therefore, mates, covariance analysis (sometes called ff the former were exluded a priori, the esti- Chow tests) permit systematic evaluation of miated impact of the latter probably would be whether subsamples should be combined, and biased upwards. such tests should be undertaken. If the tests 6.51 Scond, ordinary least squares (OLS) pro- indicate that the restriction that the coefficent 119

127 rum 63: Regression variable fw detmlmto of education Dependentvarigabls( Sdiool/cdge tten1amnce lat12 manths x C bdeenroledi toas 12 months x Private vemspublc hodcng x Bver attended sdtool x UVN ade =pee Highet dwkwacmbtsined- x H4dd ~~~~~~~~~~~~~~ HWmW _fo/wg"v& Hihet oatiu/tcin** ga~ LiteracC("ad shot tex&) x lunacy(wrieltbter)' Written csx Wduta Ever lrvatusv x Evgapprn gn' x Evo te inlgeow x Month nshr rin*tng 4 Tralidngin pimayoccpalon x Tbilng dwation In rmawy occupa x svang dum&aeom in sdoaryu opa pa Whether sat fa r xainam for vaus scd1lhg lve it post 12monAls x Pedumum on a>dmsdo _xehd x If eotrdil aiol past year, x number of daysabsnt x : CAgitive addevementest sre x EihR%%4d variables P ubudfm* hued ol wesu (A) Hua sses motwschon x Fathr' seholitm x Oer oorsmt auts Mohrmnhousehold x x ath in household Oth adtoresdents x x Ph_yseasset Roonsperocupnt x Typeo*dwellng x -oa slatus x dngs wo x C Pdftener x Gd x Toilet type x * utdwalsmaburla x FlormateraP x x emates acmss such subsampks are equated is ity bias results when there is a seectivity rle for not rejected, then te televat subsample.s d a subsample that Is based on unobsved characbe cned. esc that play a role in the iaton for 653 Fourth, thereisanissueaboutexaclywhat which the sdsam le is used but are not conubsampl should be used for a parcula esti- tboled in that subsample. 2 The is enwmical mate in order to avoid selectity ba. Selectiv- eldenc that selecvity buas may case subs IN

128 Fame 6.3: ReVuson variables few detemian of educai (contued) Rloof ulum rea proaant PVrdUdnA8 (teta value aeailtral orf Oaher enerprse Matet)AA x x x FWAr, Pro and Fhlw Mat Mwuy id Twi Pric (P) FPso markets LOW waerates (me* femaeyu Expected wagerate memees foe rekevat level ofe,dwoolng )A4for ownaterp proite x N x 31000, 0_d Marketepr ies s X 'fliecsta _tt fotes r x Smaim n/teralbn quality x :(tmea m sl x 0*"C=wwtraIingMPdlbelt b-fxbandh.wke pi Sodl Eanploymnt-latet x xtensionx 1~~~~~~~~~~~~~~~~~~~~~~~~~~ x (ef t l ch trqualficatom of tvaal s.3 I (eaehe.h x xt Trlt bettyaleatoi m n es to r*imi enth) x _Etefsi &sm x 4 fo O..w.c 25 ye a,. e.th Iamuican o a albamydpxsprtn oead ynaa io or LDaNedtbwaodvyuspulingse.Am<X oeslucaileew ~~~~~~~~~~~~~~~~~~~~~ doflvyboenalecmiuin tase dfc,s &dh Pua re on of si vlenmiay d g hs boub m u cm e sbeory LNe *- ren X X_3r- d4odm dsh ramwafw rud.dl an exh sa tpriular, VdMIVVovem oebancoheaveragedmaieur tt o b e 8 ^ -~~~~~e dnndwd SDAhusehldsrveyietamse,-- bdibltoadd Oougubcstof eopdfiaaddwevutasreuuybe tial takfgof emphricad micro tela- mulating the quzestion so tha the estimates do Lion reated to human resources in developing not depend n a seected subsample conties (eg, see the suveys tn Scultz 198b 654 In the pmeen context, the upper seocmand Behnban 1990). Selectivity bias can be dasy schoolng adm decsion providps an avoided by wrolng h for the soection rule in Iustration. Thosn who are admtted to upper the estmtes if a subsape b used 0 or by for- sewndayswhaityplayareverysdected.a They! 121

129 entered and graduated from prtmary school, they whai determdnes whether or not they enter upper entered (perhapsaftersomeexamiutionorother secondary school. 2' This may have the apparent ationing procedure) and graduated from ower disadvantage of lunming together ose who do seconday school, and they in many cases sat for not enter upper secondary school for a diverse and passed the upper seconday school admls- set of reasons (ranging from never enting prisiodns amation If one attempted to estimate mary schoo] to deciding not to enter upper sewhether or not they were admitted to upper sec- ondary school despite having passed the entre ondary school only with the subsanple of indi- examination). But it has the nmoe than offsetft viduals who graduated from lower seodndary advantage of avoiding the selectivity bias that school or who passed the upper secondary ad- would result if the subsample were Limited to nmssionsexamination(asinappletn,collierand some group such as those who sat for the upper Holl 1989 for Cte dlivoire), it would not be secondary sdhool entance exam tion. Therpossiletointerprettheestimateswithconidence fore, this general approach is recommended unbecause the failure to control for the previous less there is a possibility of controlling directly selction rules would be likely to cause selectiv- for the selectivity ries (which is anticipated to ity bias in ose estimates. be relatively rare). 655 To avoid such biases, one of two options 6.56 Fifth, there ae a number of additional isneeds to be followed. (1) The previous selection sues that ndght merit exploration fo some counrles can be controled in the esumation of the tries. (1) Interction effects may be Important, say, outcome of interest. 2' One way of controlling for between parental schooling and irastu re such nles is to include in the relation of interest a (eg., Schultz 1988ab, Behrnan 1990). Some have variable constucted from the esimated proba- argued that parental scholing effectively substibilitythateachindividualinthesubswnple would tutes for schooling nfsuctur, whie others have been selected by the selection rules; that is, have argued that it complementsuch infrastrucm the example of the previous paragraph regard- ture. Such possibilities could be explored by inguppersecondaryschooladmission,toinclude seeing what the sign of any significnt interaca variable constructed from the estimated proba- tion terms between parentl schooling and ob. bility that each individual in the subsample had served indicators of schoohng quaty are if such entered and completed primary and lower sec- terms are added to the linear form of relation ondary school and had sat for and passed suc- (U.3 (2) Exogeo incm compentb maybe excessfuily the admission exanation for upper plored instead of the assumption that all of insecondary school. However, to fouow such a come is endogenous. Typically in empirical procedure the slect mrules must be Identified. studies, for example, it is assumed that iur- Tat is, there must be vaiables that determine earned"inwome(ile.,incomeotherthanfomlabor the seection process, but which do not enter into income) is exogenous. This assumption is probthe decision of interest 21 In the case presently lematic since such income often reflects currnt being discussed, this requires variables that af- labor time use decisions inown farm/firm aclivifectdenteringandcomplen prinaryandlower ties or past labor time use decisions that led to seconday school and sitftng for and succesfuly cient returns to past savings in the form of passing the upper secondary school admissions physicaland/orfinancial assets. Nevertheless,t exams, but which do not have an impact on the may be desirable in some studies to test for the decision of wther or not to enter upper secon- endogeneity of such uneamed income by using dary schooling. It is unbelievable that it will be the standard Wu-Hausman endogeneity test. 3 possible to find variables that satisfy such condi- (3) Who controls income, not just the amount of lions except through arbitry exdusion restric- household income, is thought by some to be imtiovs in the SDA data.p (2) The question can be pornt. In particuar, there is the hypothesis that posed in a way so that the use of a selected income controlled by women is more likely to subsample is avoided. For example, rather than lead to increased investment in chid scholing esdmate whether or not an individual who su- (and i other forms of invesment in chidren) cesfuly passed the upper secondary school ex- than is income controlled by men. This may be amination enters upper secondary school from important in assessing the impact of adjustent the selecd subample of individuals who suc- policy on education if such policy affects men csduly completed that examination, one could and women differentially, as well it might due to estimate for the entire sample of 14 year olds 2 occupational segregation by sex. It is very diffi- 122

130 cult to explore such possibilities empiricaly, sible form of it) by induding the lagged dependhowever, because of the same problem just dis- ent variables in the relations lihe coefficent of cussed of the endogeneity of income; this is an theaggeddependentvaiablscanbenrpretd even clearer and more widely recognized prob- as an adjustment coefcient within the following lem if thwe incom of interest is labor income, simple model: ffefcting not only unobserved chracteristics relating to wage rate determnation but also thoee a X.1 Xg, + b (Xdo- ),1) (3) relatng to labor supply. The basic proposed eswimateswillglvesomeinsightintothisquestion where the t and t-1 subscripts refer to the time if it is possible to indlude wage rates among the period and Xd is the desired longrun level of X, pries 32 Beyond that, it may be possible to ex- ftat is deteined by observed variables as in plore the impact of nonlabor income through the relation (1) and towards which there is adjusttansfers rweceived by different household mon- ment each period to cover b of the gp betwee bers as recorded in Secton Xl of the prototype the current desired ong-run level and the previ- SDA household quesonnaire if such transfers ous period actual eve. are suffiiently common, although the endogeneity issue emains. (4) Unobeed housold, mdi- Step 3: Estting theimpat of adjustment pol vidual And commuty fixed factos may bias the on the pce and bflsndture vaiabkes that estimated results, though as dlscused above this detemine houshold educationa decins may not be such a problem for the purpose of the project since the coefficient estimates of the pa The next task is to use the reslts of pararental sdhooling variables, that appear to be par- graphs 6408 through 6.27 to make esdmtes of the ticuarly susceptible to such problems, are not of changes in the prices and infrastructural variprmay interest in themselves for this analysis. ables that enter into the right-side of relaton (1) In those cases in which the sample design leads due to the adjustment policy. As eniphasized in to a panel data, however, this could be explored paragraphs 6.08 through 627, such changes are by differnng relation (1) over time before esti- likely to be very dependent on the duration of the mating it. Such d Hernng pemits control for adjustment polcy, and may dhange over time in all individual, household, and community fec- magnitude and possibly even in sign. Therefor, tors that remain fixed over time. (5) The time it will be desirable to have sets of estimates of the period may be relevant, if there are adjustment impact of such changes for different time perilagsinvariousdecldons. Asisemphasied above, ods. the first year of the adjustment policy, a any given estmates are for a particular time pe- medium-term perspective of several years and a riod, so that the estiates may change over time. long-run perspective. Of course the longer the Generally, the available data do not permit much time perspective, the more speculative are such exploration of this possibility at the household estmates. But it is important to keep in mind reduced-form timation stage, though there is these different perspectives in order to assess some possibmity of dealing with it at the next fully the impact of the adjustment policy. For stage of the analysis in which relation (2) is used. example, suppose that the immediate impact of Given the nature of the available data, it would the adjustment policy is to reduce economic opseem that at the regression estimation stage the lions and the quality and the quantity of the only question is what assets are assumed to be infrascture due to budgetary stringency, but fixed in A. The longer the time period, the mowe that in the longer run it scessfully turns around such assets can be affected by savings/dissav- a stagnant econony and improves economic opings and further train and education even for tions and the inrasuctm In such a cse an adults. In some cases it may be useful to explore evaluation that considered only the short-un efto what extent the estimates of the coeffidents of fects probably would undervalue the gains fom the prices and infrastructure change f the set of the adjustment policy just as one that considered variables in A is reduced by eliminating those only the longer-rn effects would overstate the that would seem to change most rapidly (e.g., gains. _mesures of consumr and producer durables) The meso variables considered in para- In thos cases in whihthere are pand data, some graphs 6.08 though 6.27 and the meso variables further esdmates can be umnertaken to investi- used in the estimates of relation (1) that are disgate the lag struture (or at least a simple pos- cussed in this section gerly will not be identi- 123

131 cal. Thef, som links wl haw to be estab- For example, ther wm be a set of estimutes relised. Probaby it will not be possible to esti- garding ranges of possible induced dhanges in note Unkdng _ ehms for such ls snh the school andance for pimay-age boys for the necessary data are not likely to be avavbe. M first year of the adjustmet policy, for the me- Thezefoe, the researces will have to use thir dium run, and for the long n.m pdgemn to detnnin to what eet the mote aggregate chage measued in paagraphs 6.08 Polcy Implcations through 6.27 affect the dis variables used in the estimation of relation (2). For ex The poicy implication that can be disample, if the real central governmental resources cussed in any specific case of course depend on on primary education deines by 10% in the first thedetabisofthatcase. But this Analysis Plan has year of an adjstment p,grani what impact is indiated an apprach to attempt to quantify the that likely to have on the swholng infrastuctu- Impact of sone adjusunent policy related choices ral variables that are used in the regression esti- on educatonal investment, and thus hopefully to utes? Cnsdrations to keep in mind in mak- inform better poic choies. In is section an ingsuch udgenwntsindudethatsomeineased attempt is made to provide some general p*ieffidendes may be induced by the rsticton in ciples on which such polcy infens nuy be resources, that there nmaybe induced shifts in the drawn and to present some appropriate illustaconosition of inputs into the educational pro- tdons. duction process by the induced changes in prices 6.61 The first geal policy priple perains and scadtiesduetoadjstmetpolicy, ihatprim to the need to place the educatonal sector in a may go down coniderably for some Iportant larger perspective. The effects of adjustment inputs (particulaly staff), and that some of the policy on education have to be assessed in tems resources for local education come from sources of the tadeoffs elsewhe If a given reduction in other than the cental govenment. All of these central governmental resources devoted to educiderations imply that the prbable dedines cation would cause a 10% reduction in lower in the inditors of local schwoling frasucture secay scwoolng enllments in the fist year that are used in relation (1) are Hikely to be less of an adjustment plan, the question of relevance than in the more ag te meaures that are is what is the trade-off in terns of other objecdiscussedinparagraphs8through627. Given tives? Inorderto lessen tws reduction to, say, 5% the inwrent uncertainties in maldng such judge- how much deterioration would one have to acments, however, it probably will be desirable to cept in child health, the balace of payments, the make a nge of esmates under altemative as- rate of inflation, or in other indicators of the sucsumptions about the extent to which the i- ess of the economy regarding short- and longervenn fadcorsuch as mentoned above mitigate run goals? Although the type of analysis laid out the impact of adjustmen poliy on the meso vari- in this Analysis Plan for education is partal equiables obseved at the household level. librium in nature, and the broader issues of tradeoff are economy-wide, the indicated analy- St* EvtiMin ofhe iwtc tofadjtt sis (together with similar analysis for other parts policy on elucatioa indits of the economy) can help to give some quantitative insight on the basis of which better policy 659 This last step is storward. Simply formulation can take place. use the estimatesof the changesin the meso vari The seond generl policy principle perables from Step 3 and of the paranmes of rda- tains to the importne of the time dimension. tion (2) from Step 2 to estinate the impact of The impacts of adjusutent policy on educaton adjustnent poicy on the range of educational and on other outcomes of interest havean imporindicator included in the dependent vrables tant time dimension to them. It would be ihap- %) of Fme 6.3 for alteative time periods and ptopriate for policy evaluation to onsider either foraltenveasmio fromparagrphss657 only the immediate Impact or only the longerthrough 658 concerning the extent to which ob- run impact Both need to be considered. Otheserved more aggregate changes from paragraphs wise the chances of inappropriate evaluatn of 608 through 627affecttheind}catorsof themeso policies increase substantily. Of course it is sewt used to estimate reation (1). For any difficulto know what the te paths are kely to given outcome, this will lead to a set of estimates. be for all of the different outcomes of interest. 124

132 The approach taken in tfis Analysi Plan illus- the onern of adjustmen targets rating to the ttesamersofattinlngsomequuattvwse overall governmental budget on the meso of what the impact on education would be over lwollginfrasucturethathouseodsface. For differet time perlods. With the use of smilar a given al educational budgetl thte Is a approaches for other sectors, te poicy pespec- a of onbinations of subsidies, user charges tive over ftme with inherent tradeoff over dme s laps, and direct provision of educato can be illuminated to give a better basis for po- services that the Mhnistry of Education can proicy. vide. Dfferet combination may have different 6,63 A trd gneral piciple is ftat tere may Implicatons for the educationof dlfferetups be considerable substitution witdin the house- in sodety. Theneal appoach outined in this hold regarding the demad for education and Analysis Pan, particulay if supplemnnted by rllated outcoms. This is one important reason more analysis of the "production" of education that changes In central overnmental resources fom the supply side, cn help iuminate what diected towards education may not result in awe some of the tadeoffs in detemintng the simple prportol chanes in education. Tlhe composition of educational policy given vwious household thatis making thweducation decdions educational goals. is operating in a mewso setfting of uldtiple markets and social and econowmc Ln, all of Notes whih ic likely to change somewhat inrespone to adjustment policy. Changes in household in- 1. ltee ote Analyis Plans provide Insinto the come due to changes in labor markets may have anadysis f heimpact of(suura adjustment onthereturas greater impact on educational investments than to educon in labor market own entepes and housedo changes in educatonal infrasbtuctu. Only bol podu ctionhm. C sthm in eexpectsereturns, in um by ving the household as making educational fi.dbacntdemandforeducationwhihhesujetof decision withn this laer context and allowing i mpr. for Ial substituin within the household 2. Such a genral dcmalo provmde pat of the back can the probable effects of changes in the educa- ommd for au of the Analys Plans Addbson, Demery. tional infiatuchre due to adjustment policy on FenI, and Round (1990) provide such a dison. Rehousehold educational decison be estimated. hod refenc ae Addison and Demery (1985), behrman Ts analysis plan gives an approach to such esti- 098, hbm and Deowa (1988b, 19SM and Dmy mation. and Addison (987a)..64 A for general principle is that aggega- 3 The Ipact of adustme an actual labor mret retion maybe very important in understanding the u^ *dudi those to educatior, awe disced in te impact of adjusent policy on education. The Analysis Plan on empnomtendenwa (chapt impact may be much different for males than for r. TemayasobeIptnretursI thes females, for priary schoolng than for sewn- seor There furt may be fimnt retn in tm t dary schooling for shooling vers trainin& for beath andnutrton which are d sed in ciaptsix and smawiholders versus labor housolds, etc. Ex- seven. amination of theaggregateeffectsmay mdss many 4. Pen some ote inpub tn te education producto of these details. For example, if aggregate pro IssLpresoffactorlnpusmghtbeeTpeded4toinarao sdcoolg attendancedeclined inthe short run by Exampls wouid incude prt Input suc a speda only 3% it would be wrong to conclude that ere ked instructors and equipment at we not availe In very were not some important deleterious effects if at est s y fo domeic soura (at ast In the srt the same time primazy schoolng attendance by run). Such ces ma liwely to be mrevat relatvey ray girls in poor nua households denemd by 15%, except fspeized advanced educo The Impad on given both disthibutional goals and the high ap- thea wet pe)of educion abroadisa related, and prbably parnt total returns to sholing for femals The frm a genra poiwt of view, more revant phenotenon. approach outlined in this paper pemdts some 5. Whate m thee first two effects for the sf-elnquantitative assess_nt of such differential ef- pio,yl they asocar overin trat fator asfor fects. employee who work In the same or H,tila actvities with 6.65 AJ geneal principle i tihat thee also both nme and price effet ons agak may be important dhoies that can be made re- 6 SeeBDehnuman(IWftoareviewaempktealosudiof garding the impact of maco changs in central the Impact of helth and nutitio o chool performn governmental educationa expenditures that are me t dewvi countes. Tee oude ma suwsve, 12S

133 i~ l 4 i..2 at,t III 3,1ie i, it,,,'j l1ii, ' " " it IJI" 1

134 Us/h age pow VrouW. infity, , 1. even hoseold choc variable and sc the reat we. macy thoug suh an Inivdua inby asumptio alvays just am dfwbyse ye arah of the relevant peew group. 29. SImilarly, for the enfire sampde of 14 yer dso, other 22. Tee od not be smutaty l bea the queso cld e expled, uch as whethe or th i relao are reduced-form demand reatio so the tight- cnpldedpimaydsch mletelow!rsecaysdo sidevwrabkles awr e rexogenufmthepont sat for upptheper seadary bol a_ a of view of the houst The maidtted variable problem is and omfy completed the upper seeondy school adless than usual becae a maor form that omitted vaiable migons examinatin.the critic point ls that the ae bias aften app y takes s to rst In overestimation of sample of 14 year olds would be used for the esmates for the ImpaCt of parental s _chool due to the faie to contrl eadc queston, no Just the oves wbo had passed some previfor unobsved parental abiities, motivation, etc. (e.g., ous selectionrl Behuman 199(>. But la the present ontext the impact of 30Q It should be noted th if relation (1) has such ineacpaental sdoling per se is ntot of nterest dince t ondy 1s dlons, relatio (2) becosm more complkated byincluding, in Imotant to indude It as a controlin rdaon), hfie reut- addition totheexplicittermstudiatedabove, X, XandX, In esimates of pr_td schooig do not enter into reatin * A X 1 wher X, and X, are the two varabla tht interact in (2) V A enters into rlon (1) additively. There it i not tion (1). toubesome if partd schooling in part i9 repretingthe 31. If the uneaned intome cdmp t ffitat is to be exeffects of sudciu n*9eved partal chaaerll; In fat he plared is the net revenue (or prof onoused d entermore that it pidcs them up the beter sne that lesm te pdes, for exp net revenue is estimatedasdspendprobabilitthattheceffidentestimatesoftheothervaiables eant on the exogous as and pries that affec the enterare biased beuse of failure to control for swuh unobseved plse producon, the redduals are calcled hm t relavariables. 2in the variant of relation (1) s estimated in whkh net 23. lhe. selectivity isues are hary ldted tod e analy- revenue to Induded insteadof the exog sproduction-side sisofeducattandedaons. 7heyarepervasiveinmuch sodo- prices and asses butin addi the estimated reidl from economie analysis. Pe*ps the most emphasized e in in the net reveue fuion is induded, and a test for the estmating the detebm tsof wae rates Siven that threi, si gificac of the coffidt estimate of the reidu net a seectivity nle pertaing to who parcipate in the labor revenue indicates whe there is a simultaety prblm fre A er exampe coneuss the selecion of type of withsuchincme. SeePittandfRosnwelg(985)forsuchan heath-ae povider gen the selection prooe reprding applcon and any recent bads sconome t_ text, stuc as who cnsders _temselves to be sid lherfoe, the other M-ddla (1988, for a detailed discs of ths ewgesty Analysis Plans also needto be amnored with these isues. test 24. Suc control may be with the Inverse of the Mills ratio 32. Of course individual wage ts will not be obseved assuggested by Heckmd orby esmating the sledio8 nde for every individual In fact the same seeidvity prblem as part of the maxirum likelihood estima See Madda roted above In lely to be very impornt In detel8g (198 or any standad eowoantrls text. what wage rates are observed. Therefoe in important to 25. See the previous note. otrl for such sdectivtyin estatlng wage rates f tby are 26. This Is an exdusuon aiterbo for identcatin. There tobe used in the analysis. areother mes of kdktirfcation, such asby fncion frm, 33. If there are unobserved fxed effects, howeve, the pe but sudh approces are ot very sadtseaory in the present vkou period actual value sobuld be treated as endogenous conext since there are no peruasive reasons for the dcdhos with Identifying In ments from other previous period exof fuctdonal forms. ogenous variabe 27. If there were panel data there woud be more posbl- 34 In cass n which the necessary data for the viables ily of finding suh Identfying variables becae some of the dscussed in paragmphs 60 tbrough 627 are avalble in selecion poeses occurred at earlier imes when cundition or m, they should be used dhecdy as Indcadiffered tor of the meso setng in the etmbnatian of raton (t). 28 Separated by sex since that in not an individual or 127

135 7 Household food security and nutrition Introduction ability of expedures an food nutrient nake an other oma of fod scrity. Th spedfi Food se ls uesrgonthagedio ad estimati of rlationhips wi aow of nmy Afridcountes enggd in poliw re thealculation of key pamwet, elastctis d forl Ths Analsis Plan pvdes a famework oe fa s fthat Indicate how cha*es in te for anaig the Impact of stuctal adsmt Id variables abect _e outomes. onhousehl food sety and nut on. It also parameters can be used to indicate the seeksto acqut researchs witlh th tools re- ft of past polciand to simulate theimpact qured cy out such an analysis. Househild of futur ventos foodonumptionpatternsandindividualnt 7(B en thie wealth of anlt possbi-- ia are viewed as ftons of tie ft is important to undad thelimitatis ousehold and mat-levl fa opeating o the analys of food secuity and to= notbypricstand h nutr d n a pteu of stuctr adjust- Hypotheses are formulated regardgns in mea At the ational leve it nay be posible to meo-level varabes arsidng from stu ad- predi the impact of poqiyhngesonoutons stunet and to theesmato of the such as the level and var ity of natimal food rsting food security and nitonal effects are supply. As thelevelof anlysisis d-_wd, proposed. tbe relataonsbip beig examined beodme moxe 7.02 The key undt for food seurty analysis is complex At the level of individuals, there ame the hosehold (alhough the meing of individ- mary diffeent ift that may affect food inual-level food security is explored below), take and nutritional statusand It may be difficult wvbermsfornutridond st usevaluationitisthe td detect the effect of specific mac policy individiial. The purpose of analys at the chages household kv is to Idenify housds whose 7A) Theanalysspoposedbelow canbeused foodseatisatrk,toidythefacto*sat ina nurner of ways.mew include the hllowaffect food security and t atet toquantify ing the reatonhdat betw en tm The nutitoa (I) Prlktding the inpao of poiy reform on the satus of individual houshold mbers (for ex- food security of the nation and of diffeent ampl, young ddren) is simirl exined In grups requies an understnderms of hou d-leve variables. Me main ig of the houseold eeonomy, th reatiosp nscom i todeniy hango w in household between iousehokds and mrkets and the reincoms the relative primes of basic food and poses of househod to pubc policy refrm ote comnaxiutes and key househod and com- (ii) The _ssemen ox h pat Of past polq mmity,ifect affe level and vad- changes on food security. Hem emphasis is.,~~~~~~~~2

136 placed on the impact of policy changes on food- some of the time. This is trmed hrnsitgy food insecwue target groups. It hould be noted ftht t inseurity. Itny resltfrom regulary ocuring will be difficult to separate the impact of adjust- events, such as dry seasons, or rregular events ment policies from the underlying conditions such as droughts. The emphasis here is on the which lead to the adjustment paage. variability in food consumption and its conseir) Monitoring food security ad nutrition in a quenes A workng definition of food secwity period of policy refom. Given resource co- exdsts only when the level and time frane of the straints, this may be achieved by focusing on desred analysis is spedfied. selected household groups and indices that are sensitive to delines in the well-being of these Food security at the natillevel households. (iv) Assessing the cosfs and benefrts of policy revi Food scurty at the national evelisa major sion or of additional polides that might be intro- policy goal for almost all Sub-Sahb African duced to Improve fiod security. These indude countries. National food security maybe defined policies that may ameliate the transitory effects as a satisfactwy balance between agregate food of the adjustent proess on particular popula- demand and food supply at pdoes which do not tion groups. This is both the most important and exclude poor households from access to food. the most difficult type of analysis. Altemative definitions compare some aggrega The nder of this Chapter is divided tion of thfeoreical food needs with total supply. into three main parts. It begins with an overview While this approach has some superficial appeal, of issues facing the food secrity analyst, includ- in practice it has a number of drawbacks. First, lng the clarication of the leves and time frames the tanslation of food needs in terms of nutrients of analysis. It then develops a detailed guide to into quantities of food commodities is complex. the anaysis of hous d-level food secmity in Second, it generally ignores the role of prices and the adjustment context. The third section looks incomes as determinants of demand and the efat the analysis of the effects of structural adjust- fects of changing prices and policy refrms as 3m on the nutrition of individuals. factors influencing supply Nations are chronically food-insecure ssues In food scurty analysis when, over a series of accounting periods, food supply (which includes domestic production, 7.06 Food security can be defined in a number commec imports, food aid and changes in of different waya In this Chapter, we will work stocks) is insufficient to meet food demand at with the definition put forward by Reutlinger reasonable prices. Transitory inscurity will reand Selowsky (1976). Food seurity is thus de- sult when, from time to time, supply is inadefinedasacceesbyallpeopleataltimestoenough quate with respect to demand. Here, food prces food for an active, healthy fe. There are two wl rise and the poor wdi be forced out of the dimensions to the analysis of food security. Thle market. In general, policies related to Imprving first concrns the levl of analysis. Food security national food security tend to concentrate on fte can be exmied at the national, regional, con- stablilisationd Increase of food supply. Almost munity, household or individual level Much of all Sub-Saharan Afrian countries have national the analysis in the context of strctural adjust- pohcies designed to inrease the level of food ment will focus on the imwpt of nationd poicy supply, evpecially that component deived from reform on ousehold and on indbidals. We will the national production of basic food staples. The beparticularlynteiestedinthosehouseholdsthat addevment of national self-sufficic in basic are food-inseme and those whos food security staples (which is rarely defined priedsely) is seen is threatened by the adjustment process. The as a key national objective. The achieveamt of seond dinesion neates to the time fnme. For national food security, however, presuposes example, households may be unable to gain suf- both adequate and stable supplies (obtained ficient access to food on a continual basis. This is through an appropriate combination of domestic termed oic food nsrty. Here, analysis fo- production and trade) and suficient demand or cuses on the levd of food consumption and the purchasg power to guante adequate access factors that determine this. Altenaatively, access to food. The demand side analyis of food secato food may be adequate on average but vari- rity has to be caured out at the household level, ations result in access to food being inadequate where individual food consumption is deter- 13i

137 mined. An asof the level and the vari- food insecuisthe phnomewn of households abiity in the availaility of food from all sources, not having enough food to eat, it is not the fact of and of food needs, is equied. The lattercalls for there not being enough food. In the context of the use of nomnative measures of nutritional re- structural adjustment, the analysis of household quirements. food seurity needs to be coened with the impact of adpstment polcies on the entitlement Food seurity at hehoosld el set of different types of household and on the prices, or exchag rates that these households 7409 A cncetual fanmework for analysis at face. the household leve is provided by Sen's theory 7.12 Investation of these issues is not easy; of exchange eitlments. ' Houselolds can be there are a number of both theoretical and pacticonsidered as having a set of entitlemnts. These cal problems to be faced. Even though the SDA are defined by the economic, social, legal and Integrated Survey in most countries wil be a agr-economic environment in which the house- complex and wide ranging data collecion exerholds live. They indude entitlements to own cise, I will not be able to capture aul the subtle production, to trade, to the labor of the house- and detafied aspects of household food security. hold menmbers and to a variety of tansfers, in- For example, in an Afiican context, household cluding social secuity payments, food aid distri- entitements may include assets and incons that bution and other mehnisms. Households can households can claim from kiniolk and the state choose to use their entitlements in many differ- when income shortfalls are expedenced. These ent ways; for example, they can choose between wil include traditional links with other houseung ousehold labor to produce food and other holds which Imply transfers in times of need as agicltural commodities or they can seek em- wellasofficallyorgnised li progrms. lhese p-oyment and purdcse food from the income tnmsfers wil only be directly observable in times they obtain. What they choose to do will, in part, of crsis, but omitting them will lead to an overbe detmdned by the exchnge rate or price at statement of household food insecurity. 2 As alwhich fteentitlements can be exchaned for food ready indicated, food security has a time dimenand other goods and servces The set of exchange sion which is difficulto capture in a single crossentiments is the set of food and other commodi- secdtonal survey. While countries will be encourties which may be obtained with all possible aged to repeat the S at regular intervals, in many combinations of entitlements at given prices. places this will not be possible. It will be impor A household wfll be food-insecure when tant, theefore, for the analyst to realise the limiits set of exchange entitlements is insufficiento tationsofthedataheorsheisdealingwithandto meet the food needs of the household members, ensure that the wider ptumre is not obscured by taking into account tha# food is only one of the the details of the data that are avadale basc needs of the household. If e exchange entitlement set is insuffident to meet food needs Indidual food securnty a nutrition al of the time then the household will sufer from chronc food insecurity. This is a function 7.13 In a similar fashion, food security at the of poverty. Transitory food insecurity occurs with level of the individual can be defined as cotinua temporary decline in the exchang entitlements ing access to adequate food to live an active and sel This may result from a regularly occurring healthy life. As analysis shifts from the houseset of events such as seasonal variation, or it may hold to the individual, however, food security is be the result of factos such as drought or other affected by an increasng number of factor. Inenvironmental calamity or the impact of macro- dividual food security can be viewed as an policy chage and stural adstml individuals claim on household food resources 7.11 The exchange entitment conept implies in relation to the needs for growth, work and the that the food insecurity at the household lvel is a maitan of essential body fumctions. A "foodresult of inadequate demand for food. Total secue household enviroment Is clearly an imhousehold income, indluding the value of own portant determinant of individual food security, production, tafers and all other incme sources but it is neither a necesary nor a sufficient conis not sufficient to provide the household with dition. Animportantfactordetertninngindividenough food to meet its needs at the prices the ual food security is the relationship between household faces. To paaphrase Sen, household meers of the household. Food security at the 131

138 leve of the houold does not necessarily mean discsd furh in the next secdon below. that every Individual household nmmber wil 7.16 Within the contex of food secuity, proreceive sufficien food to meft his or her needs tin-ey malnutrition PM b the most i- Alterntively, food-inecure housholds may re- potanutitionproblemfaedbyaicanuconstain the conunpton of some members of the trie (Lathamn 1984). Oe nutrition prblems do household to prowt the vulerable Individuals, edst,however,foreanple,vaminadefidencr, or they may distribute food in accordance with iodine deficincy and anaemia. In a number of power andstatus,thrbycwroimlwsngtho countries th irddence and severity res4t in ecurty of weaker mmbr.the mechanis major public helt programs Structural adjusthat detmine how food and ote resources are ment may wel have an impact on these types of distributed between household members are malntition, but the main chnidsm will be complex and diffiult to mnasue in lare scale throughthehangesimposedontheoperationof wrveys. In SDA policy analysis, the main em- health servioes. The effects of adjustmnt on phaslswibeonthsemechanlsms thatarellkely these components of nutrition, therefore, are to be directly affected by adjsusent measue, csderd in the Health Analysis Plan. These may well indlude dhnges not only in levels of income, but also in who actually eams it. An qpvw to th anysis These issues are discussed in more detail in the Analysis Plan on the impact of adustment on 7.17 The potential uses of the analysis have women. been outined above in paragraphs 7.01 through 7.14 At the leve of the individual, much of the i order to reach the stage of policy analysis analysis will focus on nutrition and in particular of deciding what needs to be done, however, a on measesof the physcl mpact of Iodequte number of steps will need to be completed. In food consumpdon. In part this is because of the general terms, the stages can be outlined as (a) problems assodated with the accurate measure- Idenfg the nature of food isecuity in the merit of individual food and nutrient intakes, but contry; (b) specfyig who is food-insecure and good nutrition is an important objective of poliy Idenafying the nc ti of th housein its own right Household food secuity is holdsandindividualswhoswfferfromfoodinserelated to nutritional status, but it is clearly not curity; and (c) establishing the causal fact and the ondy fctor involved. For young children, on nim whih debtmine and are MPated to whom much of the analysis will concentrate, food security. Clearly withn SDA, thv. emphasis protin-erg malnutrition is related to three in this la stae will beon the specficftion of the main factors the incddenoe of disease and infec- relationships between the macro-adjustment ain child care practices; and the ablity of the measues and food insecurity at the level of the household to maintain its food secuity. The household and the individual. analysis of nutrition in the context of food secu The analysis of both household and indirity, therefore, must be closely related to the vidual food security is a demanding exerclse. analysis of health (Chapter Five). The approach adopted thoughut this Chapter 7.15 The emphasis on the nutritional status of is to undke the simpler anal steps frst, young children is jusdfled by the fact that this and then build on these using more sophisticated group of the population is most vulnerable to iu While the order of the preetan inadequate food intake. bdficent ingestion of below places some emphasis on the esmtion of calories and proteinbyyoung cildren wiresult consamption parametes the analyst is strongly in poor physical and mental development a!nd advised to start with a detailed examination of may well be a mjor cause of mortality. In many basic tables and destive statistics. The me- African countries, malnutrition Is thought to be a chancal appication of econometric techniques, major cause, or an finportant contributory factor without a thorough undertading of the data in many infant and young child deaths In addi- and the dtuation they are describing can easily tion, the indication of the malnutritlon through leadtonisleadingrultsandwrongdnkduslons. the anthropometric t is a well developed technique that can be relatively easily ap- Te analyis of household food seudty plied in household urveys. Measues of the nutritional status of adults can be developed, 7.19 As housholds operate in an economic however, and the use of ese techniques are environment detemnned, at least partially, by 132

139 public policy, ts secton is divided into two 7.22 Empiricalandtheoreticalworkonthe(rupars Below, linkages are developed between a) hould eonomy hasbeen sumnunrwd m adjustment polici household economy and a volhme edited by Singh, Squire and Strauss food security. The discssion then tus to ana- (1986). The authors develop a feasile way to lyticnwtalm ods for analyig housedod food se- model househod responss to polcy changes cuity and nutrition. The focus is, in paricular, likluy to affect food sew ity. Specifically, the on the esimaton of income-food expditure appoach allows the analyst to icorpote both elaicities, thee ationof demand ystemsand production and conmption behavior in the the identificadon of trget groups for food policy same framework. interventions The approach bysingh etal (1986a) can be summarised as follows. Households are assumed The househo" exnwmy, adustment and food to have a utility fction conisig of three arguseurity ments: an agricultual staple;a market pwurad good; and leisu This utilty function is maxi- 7.2D The defirntion of food security outlined in mised subject to a cash, tme and Wehnology paragraphs 7.06 through 7.18 indicates that consraint. Consequently, households nake dehousiolds bewome food-insecue when the ac- cisions in dee areas: production; consumption; quisition of food faus below that required for all and labor supply. With respect to production, i members to live -active and heathy" lives. Ac- is assumed that households wfll seek to maxtcodngly, for a household to be 'food-secure, mise profits by setftng man coss equal to the following balance suld hold: marginal returns. Provided the household knows the exising wage rate, ft can determine the opti- Ac 2R mal amunt of labor input. It does not need to decide how much food It wil conme nor how whereac =B+O+G.A+&%St much labor it will supply itelf. Thus, production decisions can be made independently of Ac is household acquisition of food expressed consumton and labor supply. However, conin nutrient units; sumption and labor supply decisions are not in- R is the sum of individual nutrient require- dependentoproductiondedsions Thesedepend ments; on both prices and inces. Although pricescan B is food bought be taken as given, incomes are partly affected by 0 is household food production reained for production decisions. Thus, consumption deciconsumpto ; slonsarenothineendentof productiondecisions. G is food received as private gifts; Thisistemned the profit effect and is incorporated A is food aid reeived as wages orgifs; into analyses of the effect of price changes on WSt is the net change i ehold food sto. consumtion. The fonnua derivation of this is summarised in Box 9.1 (Chapter Nine). 721 This food balace equation ig ts the 724 It should be noted that these models asfact ftat househokls use theirenttemen to ac sme a perfectly competitive makt for labor quire food in a number of ways. Italso highlights existsinrwlarems InmanypartsofSutb-Saharan the fact that they may be both produers and Africa, this is not the case. Furthrme, they consumeti. It is Important to be aware of this ipore the fact that production and ependitures feature when assessi the impact of structurl may be differntiated by gender. It is also diffiadjustment on household food secmity. For ex- cult to incorporate risk and seonality effects. ample, raisig food pries might worsen food Production and consuption dedsions are not security by increasingt cst of food. However, separble where health or nutrient intake affect i also increases household inume wich, ceters labor productivity. Where separity does not paribus, shoudm improve food security. Before hold, it is necessay to replace net farm profits goin on to a diussin of the impact of adjust- with input prices and quantities. ment podcs, it is usefu to ummalse the rele 725 Within the conext of the model devevant literature in this area. The analyst is also opedabove,adjstentpolicesaffecthouseold referred td t chapts Intis Volume on educi- food security by changing prices, incomes and tion and healtb, as wel as the SDA conceptual the institutional structure in which households famework (World Bank, 1990). operate. The linkages between policy shifts and 133

140 Pi-ure 7Jnks between maam plicy and ood secuity *acro Exdiangerm a&po1 I Ip Mw policies IAm I' pad4 Mlarket 11 dd s / 11\US&/- 1 CO M134I \w the ousehold eonomy awe Ilustrate in Fi with echange rate lieraizaton and fiscalcnture 7.1. As the SDA conpanion volume (Worl sftraits whil affext-the cost of importe fam in- Bank, 199) contais a detale discussion of po 1 - puts and food in both nomidnal and rea tems. icy change under stuctural adjustmnit, discus,- Liberalisatlon. of financal markets will affect the sion here Is lied to policy cnes with a d - oost an avalabi of foma sector sures of recimadc on househlwd food Security. It shoud CapitaL 3lab" orarets, partculary in urban be noted that the net effects will-be specifi to areas are also likey to undergo cbakgw durin dlfrei incme roups, to ntet producesand theo adjustmet process. Fiscal austerit and in.- consmesand to different regoions particularly stitutlonal reformns ofte reduce publi sector whre mrket re poorly integrated, emloymet Wage rate r\,lations may also be 7.26 A typca outcome of adjustrmet policies chaged. This mlay lead to Incrase in umma- Is the liberalization of pmoduc ma,*ef Indluding poynmen and reductions in the income of "y food markets. The remtoval of regulations and' urbanhusehvolds. It will also havecnquce sbiisfrom domnestic wmarets will affect the for households in rual areas, where remittance relatwe prices between farm inputs an outputs constitute an importantt source of Woncoe. and betwea prodiuoe and cdnsumer prices. The 7.27 The net effect of strctura adjustrmet oni reduction In regulationof extemnal trae,together rural hqusehold entitlemtents wil depend on'

141 several fors. Incomes wlll rise if additional tratng partiuarly on thosewhose food secuity revenues derived from producer prices are not may be threatened. The methodology for doing offset by higher Input costs or reducti tn ye- so b disussed in paragraphs 7.57 to ndttuance income. Where households can switch resources from less remunerative activities to Household lel analis usig SDA data those receiving increased prices, futher gains in Income wfll accrue. Should these offset the re Household level food security anaysi duction in welfare associated with higher con- tycally involves a number of steps. In this seesumer prices, household food security will im- tin, we propose th folowing order of activities prove. However, food deficit households will be () examining the relationship between n_sur made worse off if they cannot Increase produc- of food consumptionandconomicfactkslrgdy tion and if they are unable to increase their non- operatig at the mreso level, sttng with an asagricult incomes. Finally, the benefits will be wsmient of the rlationship between food and region-specific. Food producers in remote areas nutrient intake and income, incuding the estiwill almost certainly experience a rse in input Aationofexpenditureandcalorieelasticities; (ii) costs. Output prices may fall despite an overall extending the analysis to incorporate prices and increase in nominal producer prices Food pro- the estimation of price elasticties; (iii) developducers doser to entry points and centers of de- ing the analysis to include the geal econondc mand are more liely to enjoy a relative improve- and social environment in which households live ment in their trading position. Below, the meth- through the specification and estinmtion of odology for analysing the relative strengths of household models; and (iv) usin the results of these effects is outlined. this analysis to identiy food-insecure households Entitlements of urban households wfll be The results of these activities lead to an examinaaffected by changes in incomes through the ef- tion of the policy Implications and recommendafects of structural adjustment in the labor market. tions for the development of data collection sys- They will also be affected by changes in prices, tems though this will partly depend on their previous access to food at sbsidised prices and changes in 731 ESnMATNG INCOME - Oo ExeNDmwr supply brought about by price incentives to pro- amicrrs. The discussion of household food duce and trade liberalization, security noted several conduits through which 7.29 The analyst should pay particular atten- households obtained food. In analysing the relation to two interlinked features when examining tionsip betwee income and the Consumption the effects of adjus_tmt policies on households. of food, emphasis is placed on explaining how These are the time scale and household groups. changes in the former affect the latter (ow in- Households whose food security is likely to be come changes in light of structural adjustment threatened by these changes during the transi- programs has been discussed earlier). This is tional period include rural families depending accomplished by estmating the elasticity of deon rural labor markets, the families of state sector mand for food with rspect to income. As this is employees made redundant, and low income relatively straightforward, the method of doing urban families facing dedining wage rates rela- so is outlined first. In the next section, the analytive to food prices. As the length of the "transi- sis is made more complex by extending it to intional period' is unknown, it is not appropriate dlude prices. to assume that these short-run changes will dis- 732 Existingstudiesdiffersubstantiallyintheir appear quicldy or automatically. In the long run, estimates of income-food expenditure and inadjustment policies should increase household come-calorielasticities. The former range from income in relation to food prices and therefore 0.01 to 1.18 (Bouls and Haddad), depending on improve food security. Producers should gain the conuodity inquestion, on differencesinesti. from adjustment policy reforms, particularly if mation procedures and on the extent to which they produce export crops and have access to food groups were disaggregatd. The analyst is inputs and markets for their products. This sug- referred to Bouis and Haddad (1988) and Behrnan gests that a major objective of the analysis should and Deolalikar (1988c) for a detailed summnay of betoidentfand,if possible,predict which groups these. Income.calorie elasticities appear to be of households are likely to gain and which are generally rather lower for reasons which are relelkely to lose from adjustment policies, concen- vant to SDA savey analysis. ' Recent evidence t35

142 swsts tatw hehousehod InWcme C.c (Z) houseiolds, face a dedine in the value of iheir income, they are able to ntain calorie acqulsi- where C i a me of food or caloric acqulsitio by switchng to lo quality or otwmise lion; cheaper foodstuffs, redudng waste and enter- Xlsamnmofinomeorexpendilture; taning less These "adjuntst show up in Z is a vector of other household haaclower income-calorielascities - that is, when teristics. Incomr falls calorie acqisition falls less thn food expenditure. Finaly, Strauss and Ihomas (1989) 7.36 Prior to estimating ths model, it is neceshave suggested that the calorie-expnditure curve sary to specify a fuctoral form for this relationis kinked. Their Brian data suggests a posi- ship. There are a number of possblities and a tive relationship between nutrient Intake and ex- useful theoretical review is available in Deaton pendituresforpoorhouseholds. Thecurve inks and Muellbauer (1980) and Mlundell (1988). A between 2500 and 3000 per capita calories per fumctional form suggestd by Deaton (1987b) is: day and is flatter at higher expenditure levels The first stage in an examination of the w, =a+b ln(x/n) + cln(n) + E (diz2) + u, relationship between income and caloric acquisition should be a descriptive analysis of the data. whewe: w, is the budget share of the ith covnod- This includes the calculation of means, standard ity (-(pi q)/x ); deviations, coefficient of variations, maxdmum p,is the price of good i; and minimum values. Doing so identifies poten- q, is the quantity of good i purchased; tal problems and erors and indicates specific x is total household per capita expendihypotheses for further investigation. Also, dif- ture fent derived variables can be assessed for their n is household size; potential use in the analysis. Z is a vector of other relevant variables; 7.34 Frequency distributions and cross-tabula- u, is the error term, and lions are also useful at the preliminary stage. But a. b, c, dj are parameterstobeestimated. it should be noted that correlation between two variables does not imply causation. Also, vai- 737 Although there are alternative functional ables chosen for a cross-tabulation may be corre- fotms, such as the semi-logarithmic (pi q, - a + b lated with vaiables not included. A high degtee n(x)) and log reciprocal (ln(p 1 q 1 ) = a - b x-' ), use of correlation between two variables may actu- of the specification outlined above is preferable. ally refect this omitted variable. A sanple crss- An important ixquirement of any system of Engel tabulation is shown below in Box 7.1. curves (of which this is an example) is that they 7.35 The next stage is the econometric estima- satisfy the adding up p1perty. That is, if they tlon of the caloric acquisition-income relation- are estimated for all goods, the predicted budget ship. In general terms, this can be expressed as: sbhares will sum to unity. Hlre, this is satisfied where the following conditions hold (Deaton and Box7.1: Coss-tabulation of tot food expenditure by expenditre g p Case, 1987): Ea- 1and b,=o Telfoe Finally, it is easy to compute elasticities. The _WI T ap formula is: I DnI )w 2 29S W 29A A2 where: w, is evaluated at the sample mean. 5 9a9 a61 An 47S US 738 Te dependent varlae household food expenditre, can be measured in avariety of ways. :&W T ibbm3.tdsmdhabwbdduffsq9. 1ods Idealy, Individual csumptbon of food shold Affi=W1Wpm8Una _ Tt*pmaitgjib be measd in order to compute caloric and protein intake as this minimises neasurement 136

143 error and hence estimator bias. 5The main prob- In bulk on an infrequent basis. Houseolds re lem is in bmaurg this vaiable. Accurate esti- porting no expenditures may purchae food, but mwaton reqres the weghin of food as it i not duing the refe period of the queston actualy eaten which irplies a considerable de- naire. gree of intrusion by the enumerator. In additdon, Tht wnit wof maswww - Food acquisition can be this requires a great deal of time, highly skilled ressed in either per capita value terms, as a enumerators and a high degree of cooperation by share of the hhold budget or in nutritional respondents. Altematively, it can be obtned on units. It may be necessay to value conmwdities a recal basis by interiewing the Individual re- that have not passed hrough the market and do sponsible for prepring and distribufting food. not have a price assocated with them. One pose- The drawback to doing so is that it may greatly sibility is the corwnmer price prevailing in the increase the time requred to complete the inter- locality at the time of the transaction. This may views. These costs are such that neither method be obined from the ommunity questionnaire, has been included in the IS questionnaire. or possibly by averaging ovepries recorded for 7.39 Alternatively, food cosumption can be the samecomodities that have passed through assessed in terms of food acquisition. This in- the makel However, it should be noted that dudes consumption from own production, pur- different households inthe sam comunity may chases, payment for labor in food (as opposed to face different prices. For exame, poorer housecash), net food tranfers td individuals not resi- holds, who purkhase food in small quantities, dentin the household, food consumed away from may pay a higher unit price than wealthier the household and free (or subsidised) disxtu- households purchasing lager amounts. Extion by the governmenthrough food aid projects pressing food acquisition in nutritional tms reand programs (for example, school meal pro- quires data on the nutrient composition of food gratns,food-for-workprojects)andgifs. It should iters. Both WHO and FAO have publshed such be noted that food acquisition is not exactly data for the main commodities consumed in Subequivalent to food consumption. Diffene be- Saharan Africa. Commodities should be disagtween them can be ascribed tod (a) changes in gregated as far as possible prior to calculating food swks held in the household; (b) waste in caloric and nutient values. Failure to do so may preparation and after eating (c) the net differ- generate misla3di estimates of the impact of ence between the amount of food provided to changes in independent variables on food acquinon-household mebers and the amount pro- sition. Behrnan (1990) and Behrmn and Wolfe vkded to members of this household by other (1984a) suggest that nutrient-incom elasticities households; and (d)changesin food composition may be overestimated when fixed conversion arising from preparation and cooldng. factors are applied to aggrgate food groups. 7.A0 To measure food acquisition, a number of 7.41 A key independent variable issone measfactors need to be specified. These include ure of inme. While it is possible to compute The time period - Food acquisition refers to a income from the IS questonnaire, its use is not specified period of time The eumweration plan recomme--nded here. There are seal reasons for for the SDA survey envisages that data on food this. Frst, inncome is not exogenous as it reflects consumption are colected during the second visit decisions regarding the amount of labor supplied to the respondent households and will refer to by the household Secondly, nutrient intake may the elapsed time since the first visit. It is in- affect labor productivity (see Strauss, 1986 and tended that this period will be two weeks, but in Benan and Deolalikar, 1989b). If this is the practice it is not ortain that It wil be the same for case, the estimated impact of incowm will be biall households. It may be necessary to normalise ased upwards. Finaly, estimates of inome may expenditure and consumption data to the same be lible to under-report by respondents and time period (14 days is probably the most appr- may rflect large short-mn volatility. priate). Accrdingly, it will be important to in- 7.2 Given the difficulties associated with the dude the dates of the first and second visits to use of household income as a regressor, it is neceach household on the data file It is imnportant essary to use a proxy variable. One possibility is to note where the visits fall with respct to the total household per capita expenditure. This agriculturl production cyle. Household food variable is less liky to have nmea me eracqusition is oftbm related to seasonal considera- ro, is subect to fee short-term variations and dons. Finally, some conmodities are purhased can be conidered eogeous. However, there 137

144 are several disadvantges in using it as a regres- on food acquisition tbrough promotig a greater sor. First, it wes the link between policy awareness of the imrac of good nutito chages, household ircome and demnud for Women's educatio is pcartulary Imorn in calie Also, it does not permit the analyst to this regard. It also proxies for greater productlivexamine the effects of control of income on ity in maket and household activities and, for }ousehold expenditures. The analyst is refenred women, increased bargning power wiin the to the Analysis Plan on Women for a discussion househld (Behman and Wolfe, 1984a). Of this issue. ' Other wvables -These inude variables such as 7.43 More recently, It has been agued (Alder- household location (urban or mral). man, 1990) that in the context of the producer- 7AS A selection from Alderman's (1990) esticonsumer model of household behavior, per cap- mesofexpenditreeasticities for Ghaa islbsted ita expenditures are a choice variable. As such, in BDox 72. Alderman uses the budget sharemode induding it as a regressor wil lead to biased outlined in paragraph 736, above. His indeestimates. This problem can be overcome by pendent variable is the predit natural logausin two-stage least squares estimaton. iw- rithm of total per capita household expenditures. tially, household per capita expenditures are re- In Box 7.3, details of the approach used by Bouis gresed on variables reflecting household demo- and Haddad are outlined. The analyst may also grap:hi composition and assets. 7 The predicted flnd it usefil to cownsult other recent studies, inper capita expenditures are then included as re- duding Behrman and Deolahkar (1987), dhisgressors in least squares esimation of the fhmc- inghe (1987), Alderman (1987), Behrman and tion oudined in paagraph However, it Wolfe (1984a) and Ward and Sanders (1980). should be noted that within the existing literature, there is no consensus on this last point. For 7.46 ESnMXmC PRcOD sxwen wnaexample, the studies by Behrman and Wolfe wes. The analysis described in the previous sec- (1984a) and Deaton (1987a) use per capita expen- tion highlights the consequec of changes in dittres, whereas Aldeman (1990) instruments household income, proxied by total expenditure, for them using the method descnbed above. The on the denand for food. This section extends the analyst may find it useful to use both methods to analysis by introducing prices. see whether they lead to significantly different 7.47 The simplest approach is to include a set results. of prices in the general model of the demand for 7,M Ohercharacteristics to be included in the food. That is analysis will depend on local requirements. Anong the variables which are conmonly in- C - co(, P, Z) duded in studies of this type are the following HousD,ldOstructure-Allotherthngsbeingequal, where, C isa measwre of food or caloric acquisilager households will tend to consume more lion; food than smaller households. In order to com- Xis ameasure of income or expenditure, pare consumption levels and composition be- P is a vector of prices; and tween households, it is necessary to account for Z is a vectdor of other household chacdifferenc in household size. This can be done terisics. in a number of ways. A simple way of doing so is to use a 'head count' of the number of members 7.48 The relationship between food conrsumpof the household. However, this does not take tion, income and prices can be expressed in a into account varatons in household composi- number of ways. The analyst is encouraged to tion. A household with six adults is liy to refer to Deaton and Muellbauer (1980) and Blunhave quite different consumption charcteristics dell(1988)foradetaileddiscussionof theproperfron one with one adult and five young chldren. ties of these models. Perhaps the simplest func- An alternative procedure is to weight individu- tional form is the double logarithmic, that is als in the household using a weighting factor to calculate the number of adult equivalents or the ln(q) =ai+ b 1 ln(x) + Z (c ln(p?) + Z: (dqzj number of consumption units. It may also be useful to indlude the age and sex of the house- where- q is the quantity of food i consumed (see hold head. paragraphs 731 through 7.45); Education lels - Education will have an impact xis the value of total household per cap- 138

145 I~~~~~~~~~~~~~ Er~~~ ~ ~ J

146 C -c 0(, PZ,7Y*) Box 7A Esimati price elastiities where: Y* s household ful Iname as outie i Box 9.1 (ChapterNIne). Her two 9st of lastcite weprmnted Set (a)r 1mm fsagh SquIre and Strain (196. Owoptr 1) and 7.52 Housshold fu1 iotk reflects two swom ware derived using the tmatkm poedure dezpbed mabve. Set (b) are om Stn (1986). Tlbe differ in ponents, net fanbm profits and the tot value of twoy ausaquaofy household tme. Details on calculating net farm n i abow the dope of the agut airue to vary over profits are found in the capter on the Small- diffarentiname &gr. Serzd be ainvert acqilholder Sector. Tine can be vahled at the prevail- tio of food ino alozic avalabity prior to eslima ing wage in the custer or by esdmating wage prieedty. equations using data on individual characris- e tics. The Analysis Plan on Employment and Eaingp contains a discussion of ths point.' 'Wd dice 7.53 The nex step t is to specify a functional CM5ft "C 1 c pwv sbup form. A relatively straightforward posbty is _s -a.. thellnearexpendituresystem(ls). Thishas the $amlem Rim 474 Q6 advantage of using expeditumes as the depend- MNeda S&reiw -OM 0.19 ent variable. Consequendy, t avoids the diffcul- Set(b) ties noted above that are associated with dividing expenditures by dusterpies. For example, If there are two foods whose demand parmnetes C D* r p9fib Pr are to be estimated, the LES would look like the l6iafrnd vice Low fodowir; Maid& -cm 4 food - 1q,0p,(a,+t,)+w, 1[Y-p,(a+t) - P 2 (a, + t,)l food 2 p 2 q2 - p, (a2 + t2) + w p 1 (as + t 1 ) -p 2 (a6 + t)l An -a8 425 where: ;, t are taken as linear functions of hold producer-consumermodel. For househokls household demorphic characteristics; and al- that are producers of food, estimates of prica6 are parameters to be estimated. elasticties of demand that omit the profit effect wid be overstated. Indeed, inclusion of the profit 7.54 The LES equations are estimated using effect may cause demand-price elasticities to be least squares regression. If there are many zero positive! Ilustration of this is provided below in entes for a dependent variable, indicating non- Box 7A. purchase of an item, then least squaes wlll pro 756 In this discussion, the dependent variable duce biased parameter esimates. This difficulty has been the level of expendtures on food items. can be overconm using a Tobit esftmation proe It is possible to express these in nutriional tems. dure. Strauss (1984) provides an example of this. Paragraphs 731 through 7.45 explain how this Is 755 The prce elasticity of the demand for food achieved. This is useful wher the policy ques- 1 is calculated by differentiatig the LES with ffon relates to the acquisition of calories rather respect to the price of food 1. As the exact funk- tan expenditures on food. tional form is partly dependent on the maner in which fann profits are calculated, the generl 757 THE ImmEw1cA1 op 1m4mECu HOUseform is preseted hlre: HoWs. A basic aim of the analsis of food security and nutrition Is to identity who are food- (dql/dp 1 ) - (Sq,/Sp 1 ) + (86q,/6) (8r/8p,) insecureand whattheprevalenceof malnutrition is nd where t ocum This s a key step in The first term on the right hand side captues the tanslating the results of the analysis into action, ordinay effect of primes on demand and will be both for the review and development of policy as negative. The seoad term captures the profit well as for the development of more refined data effect. This Iustrates a key feature of the house- colection sytems. In many counties the aim 140

147 willbetoesbishmonisystesthrougha hous*hoks IdeBfied as potentially foodinseseries of Priority Surveys repeated at regular cum. A key distncin is between ftwe houseintervals. With mndted resources learly It will holds eostly rural) that provide most of their be Importnt to establish whh are the key vari- basic food from own production nd possbly ables for monitoring food secuity and nutrition, ev sell surpluses - the net produces - and 7.58 Food sewuity is closely interrelated to those who purcase most of their food from the poverty. Not only are the poor ikely to be market - the net puthaise hi important for chrnicaly food-insecure, they are also at much poliy purposes to know if most of the foodgreater risk of transitory low levels of food n- insecure are producers or puhses as there are sump The sarting point for the analysis of diet implicatlons for pricing polcy. In many food security, therefore, will be the analysis of counties, the assumption is that urban housepoverty and, in particular, with the poverty pro- holds are net purchasers and nual households fle whdch is descibed in the Poverty Analysis are netproduces Analysis of the evidence fom PML Although comparions with the non-poor a number of Aficn eduntries, however, suggests will be important, much of the rest of the analysis that many of the rual poor are geneally depenin this section is concerned with examining the dent on the market for their food supplies food cnmption cr it of households ntlreupanders,1989)andthatprieriseswfor below the poverty datum lie and identifyig basic staples may wel be working to the disadwhich variables are associated with the symp- vantage of the food-insecure, at least in the short toms of food insecurity. term. 759 An important first step is to examine 7.62 Relative prices wi be much more imporhousehold expenditure pattems and to classify tent for the poor than for the mote well-off. The households by the budget share they devote to very fact that consumers are poor and that their food comnmodities. Households who devote a food security is threatened means that they have substantialproportm of their budget to food are to be responsive to changes in incomes and relalikely to be food-incure. As a rough rule of tive prices Indeed, Timmer at al (1983) suggest thumb, households spending more than 70 per that Uf there is not evidence of absolutely high cetof theirexpenditures(includingtheestimated own and crosprice elasticities for basic staples value of own production consumed by the among the poor, then there nuy be a relatively household) on food can be expected to be food- low incidecde of food insecurity. It wil be iminsecure. More deiled analysis can be carried portant here, however, to distinguish between outbybookingathefood omodities onsmed net producers and ne food consumers. by households. Bemnets law sates that the 7.63 The next step is to compare the food secuproportion of calries derived from starchy staples rity clfication with certain identifiable house decines with Increasing incomes. Food-insecure hold chaistics through the use of crossbouseholds are likely to spend a greater propor- tabulations. Relevant household chaac ton of their Income on basic staples than those inude: with higher incomes. * the leve of household income; 7.60 The estimation of income elastidties for * the stucture of income and the degree of food commodities also asslsts in the proces of diversification; identification of the food-insecur It is expected * household demographic structure (ages and that food-insecure households will have a high number of males and females); propensity to spend extra income on food. They * the sex of the head of the household; are also likewy to consume those coommdities * asset heldings, eg., total assets or the holdthat ame not preferred by the better-off. It is hugmofodnsumerdurables; important to identify which foods are inferior * geogaphical location. goods and at what level of income If there are 764 Theaimistoidentify cacteristiawhich commodities that are largely eaten by the poor, are correlated with food insecurity and which which are important in providing calories and can be used for targetng interventions. For exwhere expenditure decines as icome Increase, ample, `concentration" diagrams can be conthe these could wedl be used to provide geted structed which allow policy makers to focus on food subsides (Timmer et al, 1983). areas which might be examined more dosely or 7.61 An Imporant net stepistolook at sources madethetargetofareabasedinterventions, Oer of food, particularly for the poor and those fatorsfortaurgetngcould incudeem leed 141

148 households or households with a spedfic demo- pron of household tems of trade with re graphic structe. spect to food such as an index of food pries compared with wage rates or product prices uc&w=ns FM rou. An inportnt ap- Tis can be constructed by compwring the cost of plicaon of this analysis is in the design of pro- a basket of food commodities and the value of grams that enhance or presee household food household income. Tlie key variables to be obsecurity during adjustent. 10 These indude served are producer prices, wage rates and con- * indirect income tsfers in the form of gen- suner food prices. All are included in both the eral or conumodity-specific food subsidies; integrated and community questionnaires. * use of public enployment scetmes; and 7.69 This approach is espedally valuable if it * use of direct income trafers (in the form of can be repeated at regular intervals. This allows food aid, cash, or food stamps or ration cards changes in household tenns of trade to be montenrt the holder to free or subsidized food) to tored. With respect to produc prke, it is importargeted groups. tant to observe both interannual and intersea- A useful classification of food and nutrition sonal variations in al the major producing area interventions is provided by lmmer et al (1983) markets and for all main crops. It is also neceswho distinguish between targeted and non-tar- sary to observe prices offered by different mrgeted and food and nutrition interventions. Nu- kefting agents together with credit ar ements trition interventions are considered to be those that may affect the prices offered. Any delay in thataffectthenutrtilonalstatusofthetargetgmup payment after purchase should be used to diswithout involving an ovall increase in food in- count the price offered. Finally, nomia producer take. Food interventib on the other hand, in- pricemovements should take into account prices dude the possibty of increasing food intake of inputs (material inputs and labor prinwily) among the food-insecum and changes in exchange rates. From this, it 7.66 The estimates of income and price elastici- should be possible to calculate indices representties for the target grups can be used to deter- ingproducer termsof trade. This will providean mine the effects of different nterventions. In importantindicatorofthesunofdiffeienteffects addition, it maybepossible to identify commodi- on producers. tes suitable for price subsidisation because they 7.70 Withrespecttowagerates,som-echoicewin are self-tagetg in the sense that are mostly need to be made about the most important wages consumed by the poor and food-insecure. to nwnitor. Those most accessible and nmst sig- 7.7 Fortheseinterventionstobeeffective,they nificant include numum wage rates, agriculturl require criteria for identfying target households wages and unskdlled urban wage rates. that are amenable to administrative selection 7.71 Wages and producer prices should be prcesses. One advantage of the analysis out- compared with cost of living indices. It is unlned hereis that It establshes Unksbetween risks likely that fully calculated cost of living indices to food nsecurity and easily identifiable house- willbeabletodeeppacewithcontinuousdevaluhold chararcs which can be used as selec- ations and changes in the prices of consumer tion criteria For example, dwelling structure can goods, much less be available regularly for the beused as a wealth proxy; and It canbe rlated to consumption basket of different groups of levels of food insecurity. households whos conuption patterns are sig- 7.8 The Identification of correlations between nificantly diffrent. Nevertheless,producerprices household c cisticsand food inserity has and wage rates need to be compared with conone futher important practical use. This is in the sumerprices in order to exress tem inreal temn monitoring of changes in household food secu- An alternative, albeit partil, is to make comparirityasadjustment programsare implemented. It sons with food prices (and possibly additional is kely to be too costly to make repeated obser- important non-food items) weighted according vations of household food consumption. The to the composition of consmption by low inbest chan of establishing a feasible monitoring come ural and urban groups. Where food purstrategy is to focus on changes in the economic chases account for the largest share of total exenvironment in which vulnerable households penditures, an index of 'food terms of trade" operate -notably their access to different markets may be an aceptable approximaton. If costs of and the pries they face. Indeed, perhaps the accommodation and trasport (in urban areas) most cucial monitoring indicator is some ex- and a few Important mnanufactured items are 142

149 added, a fair indication of tends in puhasing populton. power of poor fmlies uay be obtained. 775 Tables of weigts, heights and lengths for boys and grls of stated ages in a reference popu- Analysis at the Individual level lation and of weights for stated heights and lengts have been publshed by WHO (1983). Aesssdng the natre and extent of malnutrition The ost widely used rfrence popuatiort is that observed by the National Child Center for 7.72 The purpose of this analysis is to identify Health Statks (NCHS) refewed to as the NCHS those social and economnic factors that affect indi- stlandard. vidual food security. These will be largely deter Height for age is considered to be a measmined by the environment provided by the ure of nutritionda history. A child who is short household. However, as noted in the introduc- for hds or her age i coansidered to be stunted. tion, food security at the household level (as Mindful of the caveats outlined below, height for measured by a satisfactory balance between ac- age can be consdered a proxy for long-berm food quisition and needs) does not guarantee reliable security. If a child is thin given his or her height, and adequate Individual access to food. This he or she is caidered to be wasted. Weight for Secbton outlheshow food serity canbeanalysed height is a measure of wasting and is thought to at the individual level. reflecthe current nutritioal status. Height for 7.73 The most satisfactory nwasure of individ- age and weight for height, therefor, measure ual food security is individual food intake, but diffrent aspects of nutritional status and can be this is not included in the IS surv (see para- considered to be more or less independent. graph 738). Instead, anthropometric measures Weight for age is a conposite _msue which of children aged between 3 and 60 months have incrporates aspects of both stunting and wastbeen rmmendedasproxies. Studies conducted ing. Analysis of SDA data will benefit the most in different parts of the world indicate that well from usin height for age and weight for height nourished children tend to grow at the same rate indices. evervywhere. Certainly, it has been shown that 7.77 For each child, measures of height for age well noudshed children in Afica grow at about and weight for height can be expressed in three the same rate as their counterparts in developed ways: countrie U12 Children who exhibt poor growth, * as percentages of the median of the rerence maured either by their height or their weight population; are liely to be malnourished. TM. incidence of * as a centle of the reference population; and malnutrition can be shown by comparing the * as a number of standard deviations of the growth of children in the study with a reference reference population above or below the median population of chfldren known to be well nour- (also known as az-scre). ished These measures can be compared with a 7.74 Data colected in the IS survey includes pre-determined cut-off point such as a percentmeasures of the age, sex, weight in klograms age of the median, a specifled centle or as a and height (or length for children below 24 speified number of standard deviations below months) in centimeters for all children between 3 the median. WHO recomnmeds the use of cutand 60 months that are resident at the time of the off poinds based on the number of standad desecond visit These data can be used to cmstruct viations the observation appears below the referanthropometric measures, lnluding. vnce nmedan. Sice the refece populaton is height (or length) for age, comparing the approximately normally distributed, a cutoff of heights of children with the median height of -2 standad deviations wil mean that 2.3 per cent dhildren of the same age from the reference of wer noudished chddren wil be falsely classipopulation; fted. With a cut-off of-3 standard deviations, the * weight for height (or length), comparing the proporton of false fications is orly about wtg of the ddld with the median weight of a one in a thousand. cid from the reference population with the same 7.79 In addition to the tables of refence data height referred to above, algorithms have been develop- * weght for age, calculati the weight of a lngdesithrieslopesofnonmalgrowthcarves. chid as a perentage of the median of the weights These have been written into software packages of children of the same age from the reference for the cakulation of antopomic indices, as 143

150 percentiles, percent of median values and SD agricultural activites To some extent, ths may scoes. be made up for by cue provided by cider sib In assesn tde nature and exdent of mal- lings (Sahn, 199O). nutrition, it is helphfl to employ te model of 7.83 FPnaUy,nkg.staths wilalso e- household behavior introduced in paragraphs flect parenl prefee Again, the ae two 7.20 to Here, the household ma,dmlses a isses to consider. It may be that paents seek to utility function that includes the consmption of equalise an ke easures of their chilgoods, leisure, food and health This is mxi- dren Altematively, they my value additonal mised subject to an income and time constraint, ovets equally, idependent of their disas in paragraphs 7.20 to 7.29, and a health pro- tributlonaongthelroffspring. Behrmam(1968) duction function. The health production funh- provides a useful methodology for exadning ton speiesthat an individual's health isa func- this issue. Related to this is the Issue of sex bias. tion of the consmptio of nutrients, health in- Ithasbeenagued,forexamplebySen(1984)and puts such as inocuations and preventive care, Sen and Sengpta (1983) that in some regions of household resources that affect all members (for south Asia, boys are faoed over grls in tde example, water supply) and the level and quality allation of nutient However, sudies in Afof child care. A fuller discussion, and a formal rica (Svedberg, 1987, Stauss, 1990), have failed derivation, can be found in the Analysis Plan on to find sidlar biases. Also, prefrences may Health. differ between adult members of fte household One advantage of using the production- In particular, it has been argued that mothers consumption model is that it suggests that the place greater weight on children's welfare. extent of growth failure should not be ascribed Behiman and Wolfe (1984a) and Horton and solely to individual food insuity. The quality Miller (1987) present evidence that is nsistent and quantity of food consmed in relation to with this view. needs forgrowth, activityand body maintenance 7.84 The foregoing indicates that anthopomis certainy one determinant of body size and etric status is a function, not ondy of food intake, mass. However, growth faltering is also a conse- but also of health, genetic background, cild care quence of impaired health. For example, chl- and parntal preferences. Accordingly, assumdren suffering from diarrhoea will obtain fewer ing that poor nutrition is a result solely of unadenutrients and calries from a given quantity of quate food security isincorr The analyst must food *th healthy children. The health environ- be ognizant of this when analysing individual ment, for emmple acess to dlean water and health food security issues using SDA data. facilities, wil also affect children's growth. In addition, getic endowments have also been Analzing ie ddtemin s of malnurition shown to be an important determinant of nutritional status (see Kennedy and Cogill, 1987) for 7.85 The simplest level of analysis of this data evidence frm Kenya and Sahn (1990) for the involves tabulr analysis of hdividual, house- Cte d'lvoire). hold and community c ist and anthro Children's anthrpotic status also re- pometric measures. The stics are deflects the manner in whih households manage scribed below. This ud indlude mean, stanchild care. Tere are two components to this dard deviations, minima and maxima. This can Fs, it reflects the abity of households to pro- be followed by cross-tabulations of _meaures of vide care. For example, Thomas, Strauss and anthropo ic stlatus with variables of interest Hendrlques (1988), Strauss (1990) and Barrera such as age, gener, household wealth and loca- (1990) have indicated that apometrc status tion. These are parladrly well-suted for preimp as mothers education increases. It is senting information to decison-makers who are argued that this reflects better knowledge re- notwel versed intheanalysisof nutrtional data. gading child care practices. Other aspects of Theyarealsousefulinsigg wheherhemlth care which have been shown to affect inentions, rather than food policy chags, anthroponetric status inmlude weardng age and will have a greater impac on anthropo Ic duration of breast feeding. Second, thme cn- stat Another comonuseforseeidngassodastraints may Impinge on the ability of nothms t tions between nutridoonal Idices and household provide care. For example, in rural areas, these c teist is to establsh selection critera for may be parcuady acute durin peiods of peak targeted Interventio However, it is important 144

151 Du 7.5 lxample of the a5al y be a fid an eedim poed 75 lpleoffe_t~~ I =a fldo l(nsor ths are 9f n_ntn 1i Haumn and Taylor (1981) and Hslao (1986) Pf~tagsbf6ildrumMuntvdta nwalhdaowd mand the Anysls Plan on Healtt Whie ts byab9ru9usvs AnperaqfeWnM pposcisiibuebmwmt av ~~~~~~~~~~d&rd erro m tdo9 pe"mdt ditec _vm% n.t of the effect of pdrce hges on nuttiona 7a_9 A second issue coner te indusion of sonmme of ndiidkua nutientke. A priori ths would appwar desirable given that the ' 3 ewas here is on te linkge between acqui- 4 i SA tionof foodand nutritond status However,ftis 6 zs la, nt possibe to inldude od aution dey AU I9 MM as a variable Frst, as noted In paragrph 738, thisisnotneasuedaspartofthebisurvey. Itb go nwtt ecesaty to prox~y for this, for exampl by using per capita hlold expndiwu on food. Second, in the context of the poduconser model of househld behavior, househol epe that the anayst be mindful of the cavea noted ditures on food are a dchice variable. As such, n pgraphs 7 to 7.85 when devopng and inluding It as a ressrw wil lead to biased prfeti se esuts. mae Thisprblemcaabeovm _ethmugh 7.86 Asamplecrosstbulationisshownbelow the twe of an intumental vrables tehue. In Box 75. Here, the perentage of ildren For example houeold e_xpditures on food surdedinrurala ooftbeted1voirieiscmss- are md on varables reglect househl tatbuted a t househ per capits expend de_ogaphic mpositin, sets and lotion. twe and mos educa. Tids is tken from 7.90 These onidationsdicate at the fol- Samn (1990, Table5). lwing two-sg stateg is approprate. In the 7.87 A seand ty of analysis hvolves de use fre stage per capita food e iues se m of econmetr techiques Animporatadva- gressed on the varables noted in the previous tige of ths approh is ttt alows dte analyst parrph asiga fixed effects esdmtion proce the cmtrol for the omitfed variable probln as- dum It may be advanageous to t fm the socdalld wft cromsindatto_ He, It b useful dependent va Ino natu logithms as toexpretedependentvarilelnactualms, this will rduce Ihe i rather n in relation to the cut-off poids de- pedent vribls wll indude thefo scribed In paragraph The irdendeyt van- * H d dc cohwmprrsldiop. This inabbes are deved fom n ngthe house- ldudes the number of mal and females by ae hold utility funtion subjecto the con t group. This Is fund h roster section of the noted in paagrph 723. Theme are, five types houehold (bauss, 19B8) individual atxtutes such asge a * l udes: wealth held and Fendlen prental -ated inluding In the form of human cpital, spefically ls education, age and height househld chrcter- of educonl a_anent of adudt msles and fethatappearinthmproductionhnction maleo (so &ound In the household roster); and (etch as land and pital); vibles reflti non-human wealth, incuding the sie of land community attbute such as health faes; hold (om the section on agriture), the and the pries of nmption gaods, Inc value of household capital in the fum oi lqwd food, non-food items and ler asse and machinery (from the sectons on 738 As curently structu, the IS Queson- household savhg, agriule and own businahedoesnctconalnmsumerpmies Asdse ness actites) and the value of consumer wil be collected an a cluster bas, thre wm be durables (this is found in the secion on posses. nopricevaiatin withintduster. U*gp* ossm of consmer duables). under ese IdM will tend to ulas * Hou kloao. At the very least, t wil downward etimted stadad em. This can include urban or nra bcaes. In counts wi! 1ES

152 Bax 7A: EAIJa the de ia _ of f_ h -- k IhsI0aslsX lng a= This example is tken fran a mstpaper by AJdr tmal wseize (l99to basedan ISM d allestadin Ghana. As tds to Numb folder sblin a rabile to the daba that wm be made availe n Numberwfyo nger slinp the SDA prole ft is ptanlaly usul The appoah Dwmyvaiabsfrlocatin taklnbyaldem b cmparable to diatddrbedabove. Dummy variables to Indit tiauay, he estimates the determtnans of he log f per mother th education levd at the capita expenditur and also of nes The hnmer Is a Intacto tam betwem age andtion hmcio of Om Wowing varable Th sex of the cidd Numberol malesandfemalesbyapgroup Aldrman obtains th flowing result. Height Lfr ev Number of male and tmal bylffe of education fllsas dednu age, idictg thatrouic nulrilonk _s Value of ivestk uulatve. Mothes het has a strong and potve Value of vehdes iruee on height for age. Howver, the foekla* for Valueofotherceptal moth rs ahwation is n spicl N gende bis I Ara of lad owned in different ecological zones foun&norib the dummy vadiable for uban locatgosi Co area niat. Cbidren with younge dbin ae fcy Age of head of house Iban ighewg per apta e Vaious dumny vaiables to ndiate where the house- tre sigiffantdy IVrove heit for ae hold ived and whete the head of the hosehold was Shila retsarefoundwweightfr h ghtibused female as the dependent vaiable, though the vadiable for per The predited per capita expenditures are lnduded in ca a expaidihtues not sigificant. Ties does dth leat sqtua regessions where dddro height for not perfom as satisfly as heght fr ag. Thts b et age and wight for ht are used as dependent vaiabls entrely srp ve that weight for heiht Is a shrt- Other dpedt vible Indude: tem varb and hence lae to greae ram n uihagt- Ale of thie hild in month lios. Ageofthechldsnuared differg agro-diautic zones, it nay be useful to of family include this variable as well el is an err em, and Cluster-level food prices from the community so, a,, ; and a3are parameters to be estiquestionnaire are also Inluded, though these mated. drop out when the fixed effects estimation b used Multiplying the estmated coeffidents by 7.92 Beforedisussngtheegresrstobeused, the value of these variables for each household It is worth noting that several Imporant vigenerates predictd per capita expenditure on ables are not available fum the SDA survey. food. This is thm induded in the estimatior. of These include mote's height, wemn ages and the relationhp btween ati oettic staus past beast feeding praces. Independent vaiand individual and parental attributes. If it is ables that are availale and should be nluded assumed that there exisis a linea relationsp are betwee ant po tc atus and *he inde- * Individual charactsts. These inlude age, peendt variables, then te equation to be esti- sex, birth order and unson mated using ordinary least squaes bewomes: * Parenl dacristics. The indlude educaton of fther and mother, mothers age at bit Nu - so a, X,F aa PXHT + as)(o + e, Box76outlinesanexampleof thlstypeofanalysis. The analyst may find it useful to cosult where N,ithbenutridtonalstatusoftheithdcild other examples of this pe of analysis. Recent in famiy ; examples indlude Knudsen and Scandizw (1982), X)f is a vector of haracrstics of child i Kennedy and CogDI (1987), Sahn (1990), Dawa in fairly j (1990), Strauss (1988), Pitt ct al (1990) Thomas, PXHT is the predited per capita expen- Strauss and Hendriques(1968) and Horton(1988). diture on food; 7.93 Thus far, the role of community attrbutes Xv is a vector of parenta chreristics such as health facilities has been elud&4. As 146

153 the Analysis Pln on HIalfth conains a detaied rate e tran ondl effects of adjusmet. deson of ese variables and thir use, the 7.96 The IS quesionnae wil also colect data diussion here is lindted to an overview of the on a opometc status on child under five. econonmetrc issues Direct inlusion of conimu- This can be used as a rough proxy for Individual ntyr variables, such as acoess to drnking water food secuity, though it must be noted that other and health services, wil lead to blased standard factors, such as dhild care and acces to health ermors because they are common to all house- seims, also play an importat role. Tle analyholds within the duster. This can be overcome sis of these data will provide Information on the through the use of fixed effects estimaton, but imact of adjustment on Individual food secuths causes these variables to drop out of the rity. estimated equadon One way of getting around ths, as sugestd by Strauss (1988), is to use Notes Intewaction terms such as diste to doctors by mothes educaion Since these vary by house- 1. Sen (981). hold, the problem with bias disappeas 2. Deverux and Hay (1987 efr to thee daim as Iatent lncome.v Coacusio 3. Howeve, few s_maod produces ha vs to formal cedit marets (Cole and Lal 1986 and i is undear 7.94 The analyis p_eted in this Chapter is wht happen In infomal aedit makfets we ntest ae deigned to identify the faors that affect food are raised or aedltaueelmposedon fomalse oriarseuity at the household and individual level, to kehl quantify these relatonrhisps, and to identfy those 4. Abo see ouls and Haddad (198 sand erman (1990). at risk from food Insecurity. Structural adjust- 5. SeeBouisandHade(1988)foradetaeddisson ment programs affect the prices, Inomes and the of his pon institutional framework in which households 6Studie that I that the dbution of inome operate. Consequently, fte analysis of food secu- within the houeold affects the patten of demand for food rity is necssay for the ae t of the impat andothergoodsidudethaof Guatemna(vonraunetal of past plicy rfrms* for predictng the impact i1989djamalca(ortoandmie,97),gaa(vonbrn of futue policy dhanges, for monitoring the well and india Kulmar, 1w. ben of vulnerable households during structural 7. Tbis _pet I aso discussed hw_ in 7.85 to adjustment, and for assessing the coss and bene- 8. It should be noted that sfunctional fom i not apfits of policy revisions, including those designed proiate when the estimatio of the demand for all goods is to anmliorate the impact of adjusmt. As noted being atnepted as it vilaes the ondliton that the b,;s mu to in this Chapter, it Is necessary to specify the level zero See Deat and Mueauer (198 for a disc and time frame when undertaking such an ana1y- 9. Note e that f produc and consumption decisns are sis. not separable, t is neoessary to include the prie of food, 7.95 Data collected as part of the SDA IS ques- non-food Items, wage and aop Input pries n the estitationnaie are not espeialy suited for examinn tin of the dmad eqti he manlysis Plim on Helwth national level food security issues However, prvides a futher dicuson of thi them exists a wealth of analytcal possibilities at 10 ltesetmayoyeedenotshuldnotbeundtlycostly the household and individual level. In particu- nor d they becm pmanent welare tra at the lar, It is possible to estmate the impact of changes expimse of vigoous Inomes and employment polides. in icomes and prices on the acquisition of food. 11. Thbs is alo diused In te Analysis Plan on Employ- Comparing this against a normative m ue- want and Eamings mentofrequbemesind ateshowaouseholds 1 For exntpeseehabicit etal (1974). food senuity is affected by policy reform. Thee 1. Buse the andard devati for the rence popu. results can also be used to infonn dedsions re- tln increases with age, ti name whl sb not a8e garding the mordtoring of groups vulnerable to dep_md, is more accuae than the cut-offs of 80% for food insecurity and policies designed to amelio- wet for height and 9o% for heht for age. 147

154 ~~8 The impact of adjustment on women Introduction of other d gtions i nt for the analysis of stucual adsmt Paragraphs Twe is an Iprant dfeene betwee tough 8 discuss symptoms of genderthis Analysis Plan, on the Role of Wome, and differentd consints upon t gewration of otessuciashealth. Afocusuponheqlth maes Income. Paragraphs830through833movefrom obvious sense because, as an objecdve of devel- the sympts to Dthe costrints themselves Flopen it is of direct poliy conwern. Gender is nally, paragraphs 8.34 thrgh 8.3 explcitly reoneofmanywaysinwhichdatacanbedsaggre- late gender to a series of public policies. gted and the rationale for doing so is twofold. Thrugout, examples of how African survey irst, in eaning income, women often face differ- data have been and can be used are presented in ent constraits from men. Since structural ad- boxe jusment is largdy about changing constraints, if those facing women and men ame uficendy dif- Choosing the level of aregation ferent, is iluminating to treat the nders as disdnct groups rather than studying gender Perhaps the most fundamental step in undiffezentlated averages. Seond, women and ewdonmc analysis is the choice of aggrgtion of men often have radicaly different propenities activities and of agents. Although gender-issues to consume particular public services and so amcemdsareptionbetweenagents,ltisusebudgeta chges y can have powerhflly geer- ful to begin with activities. The new macroecodiffentiaedeffects. ThisPlanwlgiveexames nomics of stuctura adusnent usually as a of how gnder di tion can add to our minimum distinguishes between the six activiundentandingofpasticulartopicssuchashealth. ties shown in Figure 8.1 (overleaf). First, a dis- It should be stesed that gender is not a topic In tnction is made between arketed and nonitself but rather a possible tion to be marketed goods and services because ordy for bone in mind when stdying a topic. The art of the former can resources be alloated thrugh the aralysis is td keep I Ption to a mnimum price wchanism.n Second, within maketed acso that it is posible to see the wood for the tr tivities distnis are drawn between capital Sometimes gender dsag tion will not add and consumergoods, and between radablesand enough to be worthwhi However, for some nontlea Finally, within trdables, a distoics It wfll tum out to be useful and for others tinction is made between protected and unproessential. With properly set up survey data the teted (or anti-protected) activities. costofflshingtoseewhethergenderdisg However, thee is no equivalent consendon is going to be infomative is uswly very susastotheappmpriatedisaggiationof agents. low. The next section seb gender in the context The key determinant of this choice is whether we 149

155 Figurea D In if of actvitieslitlhe 8S05 nhischapteritissuggestedthatonthe ige of adjustment saoe criterion a fourth daggregatlon, by gnder,isalsooftenappropriatel ethecostadits Aj lodd so and hence the decison problem are likely to differ substantially by gender. MarkedN d 8.06 Acorollaryofthisrationaleforananalyes < (6) } which distinguihes between women and men Is that, geneally, tere is notasmall, self-conained CoWISrU $OodS Cidtd set of "womens Issues" which can be appended to an otuwise unaltered analysis. Rather, the Nontr\dabto lhdable claim Is that many standard issues in resource (4) ) alidcation become better illuminated when the analysisisdisaggregatedbygender. When added to the above thee-way dof agents we anrive at a potential disaggegaton of six ae- Non-tradable ltadable tivitiesby two locations, by three o a 1) _ form by two genders, which Is 72 classes of agent. Even this understates the poteta num- 4ed un- (or 1) ber of agents since many agents will be gged PintOCtOd ~ in multiple activities rather than being defied by a single activity. Six activities give rise to 64 differnt combinations of activities (nngfr*om doing none to doingai1 six). Clearly, the scope for are concerned primily with the processes by analyticaly meaningful disageatlons rapidly which living standards are determned, or ordy leads us into unmanageably lare numbers of in the outcome of those processes. That is, different groups. The answer to this problem is whether we are edned with production or tocheckwiththedalawbichamongthesegroups only with consu n. If the latter, then the Is sufficendy large or poor to be important from appropriate unit of observation is self-evidertly a policy perpectve. Box 8.1 gives an example the individual person. However, in Africa al- fromasunrveyof Central Province, Kenya,of how though such information is useful for the im- 64 diffen combinations of activties can be reproved tetng o services and trfers, of duced with little loss to arud 9 combinations. greater policy corn is the prcs whereby 8.07 To summarize, if we seek to undersand livingstandardsaregenrated Apartof poverty the processes which generate inome then our is due to consaints upon the germation of in- disaggration must be based upon analytically comewhlchareabletobealevlatedbyappropri- purposive distinctions as to how hinome is acatechangesinpoy.theprimarypurposeofthe qired. Since there are many such distions analyis of survey 'data is to assist in the identifi- this must be tempered by a scrutiny of thesurvl cation of such policy changes. data to see which distinctions are important This implies a radically different disag- 8D8 If, however, we seek to improve the targregatonoagents Thecriterion for d- geting of public consumption services, then our ton now beomes the decision problem of the disaggregations are differently motivated. We agent: only where dedson problems differ sig- should make those disincbons whih help decinificaty is it necessary to ditnguish between snmakers and administrators designing the agents. On this crition tee dsagptons at delivevyof services to identify tose mostin need. once suggest ftemselves an4 a commony sre used: The mostreadilyidentiflabledistlnctionsbetwee (V)bytheactivity(s)inwhichtheagtlsengaged people are age, location and gender. Age and (see Fiue 8.1); (ii) by the location of the agent location are continuous variables. that is we can (rural/urban); and (ill) by the extent of the prin- go on making finer and finer distnctons and so cipa-gent problem encounteed by the agent as the question will usually be at what point is It anorganization, viz sensible to stop. There are, of course, no such * individual person (none) problems with gender the question is simply * household (mupemtle conflict) whether the disttion is worth making As * firm hiarchy). with al dia ns, the test is empirid.l 159

156 1 70 S! ' t S ~~~~~~0,,I. 4 S t7 0 FE E { 4 twjditj1 1 l i'l0 iei i. P 1 tr0 tilrx

157 B"ax &2: rnsenft gmder dfffuemm bt centrate upon the process of inome geneation. This s moe compex. but probably mor impor- Hines$ tant taw Sureof householdsin rural or b atn on ilness dufing the past three months Tlb pmim To Identify groups of tbe populatko sett pmnelomnesa 81 This section diwsses how to use surve The table dlvides th populton Int four age groups end by ender and disnguhes beween five symp- Tanzani, l983, Symptoms of gened-diffentlated constraints data to determine whether constants upo eco- nomic decisions are likely to be sgnificantly diftamsoffllns H theree nsamgefgdr/ ferent for men and won. f hey face diffret symptnm orea ote. If we fsl tomake therage nts two phenwmena should be identifiadwsmptgod d 8toftetmeen Is idiffe4ence able, one static and the other dynamic. Fhst, febeween the bemn dl Ad8. g the brenakown 9 by ge, male-controlled resources should, as a consehowv, we find a strng and revealtggender diff- quence, be distributed over econoic activities me. Women aged 1649 are a3% mare ldy to be i very differently from nae-controled resources. Olm o. This Pggegs th hypothesis ta cid bear- This is because, faced with a conmon set of in- WSg ps a physicl Strin an women which adversely centives and differential constaints, nma zing eb s thelr general health Adding the breawkngdown agents win locate their resources differently. Wn amo 16thegende 43Se ene bdcrmnme larget. Second, when the structure of incentives change, ffer frm fv, diaea or voditng but iely to so that it is optimal to reallocate resources, the maffer from dcest-based Dlness (pehavs becue They process of supply response will differ sysftise less?). In tsis erample, the geder distinction cally. Both ese phenormu are of the essence makestheageand symptom dtintdloomomre infoma- for structural adjustnent. Adjustnut is about 1v resource movents between activities. Chang- Mlty Percentage of polation HIl by age, sex ing the Inmetive strucre is merely a necessay ad p means by which such movements are induced. Assisting the process of adjustment therefore Pam% <mu means identifying those types of agent who con- trol resources in those activities which should Spwpkp S+ Toa contrc an discovering how policy chage can." ease the impediments to the reallocation of those P resources to oher activities. If women and men PY 4.8 l.7 13 i 211 areallocated veydifferentlybewee sectors then FDV A. is a given change in fte structe of incentives will C Co Oa imply different requiements for them to be mo- PrnarAhlP bile. But thedifing constraints may also imply that women have less capacity to be mobile than P 16.? men. For example, women might be dispropor- DV SA 2.6 1S9 I tionately located in sectors which need tocontiect PDV and at the same time have less capacity than men C to rearlocate resources to sectors which should Cs P? nro ill' 2Z A expand The najor resource which women control Pbdleim*m. is their labor (although, as we will see, women DXd V l nmay often not be in full contol of their labor cbdw"m. Umtime). A simple and instructive startingpint in the gender-disaggregated analysis of survey data is to compare the allocation of male and fenale S, _ ' r AwEaxwAuj_,n,DL. snp. labor. Refeing back to our discussion of disagcdaww.omcha C,Odid- 9EWT 1M.Td4 gregation, it is meaningful to do this in sveral diffeet ways. First, we distinguish betwee scope for useful desive statistics on weu- farm work/wage, work/own, businss/famfly, being is conswieal However, since it is faily and services/healt-reated time. Box 83 shows straightforward the rest of this paper will con- thaton thsbasistierearelargedifferencinthe 152

158 allocation of time between me genders Women spend far more ti workdng on the family farm, lox 83: Gen(der difffes In the wlloca soewwbat lmoe tbme working on oher fams, tio of labor dme and far less time on their own busies and in wageemploymet They spendoverwhelmingy 7 To idey Meenm hi time almotn mre me feting water and wood. Box 8.3 in- bwenandwithe vsges tigs futher by di hing between The AM table Msliftle 12 non4eisne us of Ud fenmde heads of household and wlivr. in male- Womens peadfwm ewo fwfmi,fdiheaded households. Wives' time Is even more ing watr and wood. and eigs IL Men spend far more skewed towards the fm and fetchin water and m e off-firm wad;, in eduto and as outpaient wood, with less time being devoted to buiness The seomd tble focueon 7 of tsuses of m It and wage employment. It aso dl gishes be- fis wuioamren who amthoiwn oompreswe with tween women with and without education: edu- dfer level o d., it cpawos its cated wonmen spend substantially mormetm hi wit gk Woom who have the autnmy of beng buiness and non-agricultural wage employment homueboid heads spend mote tme in non farm employwithout spending less tme on other forms of met and les time fuidilg wae and wood. Edscated wok. Finlly, it compares boys and girls in the wm spend fwr moe lime In noo fam wae employage group 615, the prime sdhool age group. Girls mnt, but not at e o fe of aciie G spend nearly three times as long working as do Vnd tea ee t tims many hems tn tee adtviboys. This surely raises the question whether itis e detmental to their pefrmance at sdhool. ufewsp ur 8.11 The other important basis for disaggregating the allocation of work time is the macro- Tweofm AG" F=Wx eodnomlc sectors identified in Figure 8.1. Of the o six activities previously identified, in most of Af- amwaewo* ria domestic productionof tradablecapital gods Ownb Workoote"tes 4 3 i suffidently minor for the activity to be aggre- WO koefdsnufms 35 4S giated in with the rest of manufacturing as a p atar "protected tradable" (tradable capital goods are kd 2w 13I an important category in the macroeconomics of oupau-t te 8 4 xpenditure but not of production). The task for Acempmong oupent 1 5 the analyst of survey data is now to assign the I I activities reported in the survey to the five renialning analytic activities. This task will entail a APGew series of choices which wil differ between coun- FEUelU 645 ties to somw extent. However, in much of Africa ca as a first approximation the activities can be fue- If p-d proxied as set out in Frame 8.1 (overleaf). k p 8.12 Of these, food production is the most w M as problematic to allocate since it can potentially be Own fr 8M part of either the protected or unprotected trad- Off4amwe wek S8 Owabualum 105 able sectorinstead 37 IS of being 27 nontradable. This isa Workoesit ccal decision for the investigator women are Weakoamoli farm S heavily specialized in food production and so 40dt UE07 S2U ES Ftbawood IS 47 whether, overall, women are disproportionately T in the protected tradables, unprotectd tradables. or nontadables sectors largey depends upon it. ba_ P UI= Antt_ P.CaIaukdjW4,.cumAn5caumoutap~o*stofigs9dam~. _AWDLBav4 The designation as between tradable and UNS 1w.a"_,o a. nontradable depends upon trnsport costs often these are suffidently high for broad self-sufficiency to be cost-minindng in a norul year, in unprmtcted on fte basis of the relationship of widch case food is appropriately garded as the producer pri to te world price. If food is nontadable. If food is traded internationally, exported then the quedton is whether the pr. then it can be classified between protected and ducer price is broadly equal to the wodd price 18

159 Fnme & Relating secms to activities dentified In surves COSMe oods Tradable Protected Unpced Man g Exprtpagrcp ultre Nonhtadabe Food a8rue Captal Od1 Nontadable C on Non-marketed goods and servces Publc sector eployient minus domesdc transport csts. If the producer ity of the principal-agent problem is whethe the price is substanially above this then food is a household is male- or female-headed. Whedter proectedtradable. If foodisimprted, then (since tis disaggregatn of household alloction is at the margin domestic tasport costs have to be worthwhle probably varies considerably within paid on imported food) the investigator should Africa. In parts of West Afria very few houseselect among the locations supplied by imported holds are reported as being female-haded, in food that which has the parts of East Africa around 40% of households highest transport costs. The key compaison is arefenaleheadedandthedistlnct ionisextremely then between the actual producer price of food hmportant and that which tether with the cost of trans Box 8A tels us both where the laborforce portng It to this loation would equal the cost of is by gender and the relations of production by imported food including its transport costs. If which it is twhe. We might find, f6or example the two prices are broadly equal, then food is * that women are skewed towards the unprotected. If the actual producer price is sub- nontradable sector (where they predominate); stn the higher of the two, then food is a * that women are skewed away from market protected tradable. Having made these assign- alloation of labor; and ments, a further feasible and important disaggre- * that within household allocation, femalegation is the contractual mode by which labor is headed household labor is more skewed towards allocated to the sector. namely through the labor nontradables than that of male-headed housemarket, through the household, or through the holds. idividual 8.16 What might these findings suggest? &13 Two descriptive statistics are useful: the Struchral adjustnent involves a reallocation of percentage of the female laborforce engaged in resources from the nontradable to the tradable each sector (by contractual mode), and the per- sector. If womenpredominateinthenontradable centage of the total laborfince in each such cate- sector this tels us that resource reallocation will gory which is female. Such an arrangement of primarily concern females. If, additionally, survey data is portrayed in Box 8A which shows womens labor is allocaed by the household, thds both an "idealr disaggregtion and a more lim- tells us that we cannot rely upon the labor maket ited example of what has been done on national to achieve this reallocation. If, within household Kenyan data. he Kenyan example demonstrates alocaon, femae-headed households are skewed that there are large gender diferenc in labor towards nonradables, and if such households allocation between the sectors derived from are numerous, this suggests that the gender of maroeonomic analysis. The protected tradables the household head may currently be a signifisector is ovewhemingly eale as i tthe non-food cant influence upon accessibility to tradable acnontadables sector. The food sector and small- tivties. farm export agriculture are predominntly fe- 817 AusefulsupplerenttoBox8Aisthemean male. wage rate in each sector (Orame 82). The wage 8.14 In practice it is not usually possible to dis- rate is some guide to the marginal product of tinguish between labor whose alocation is labor. Hence, pronounced differences in wage household-controlled and that which is indi- rates between sectors indicate either differential vidually controled. However, at least as far as endowments of human capital or differential gender is concened, a good proxy for the sever- baniers to entry. This is of interest for gender 154

160 ri:me 82 Me wage rts and the dltbu. 80K 8- The dstribution of t'he labo force t{on of t hbor force by gender by nymeaigf sedo Ih pjnkm To intate the dffeitg gpnder compou- SedrM d drw 1mw of th sectis Identified in the _a.o.... analysts, Proteded tradable. Sno this work woud be nww, te is no good ex- prteedadables imfl Itwh h tollustrateit.hece, *ansi'el Fres- Nontradableamusumer goods entatonfsurveydatas stebelow, Molowed by N d captl goods - a more UoDmid example for Kenya. Public ervis The Kenyan example uflsed both houwhold surtvy Total dda and Employment and EAn Survey dala. The X pe end of femfi I lat, beg an sb tmd suvey, misses in- Y p fh f*mal and sbsen acvities but Is kely to be more ait 11'fr hnv I wa8e emaomt. Note the verykwdfeue_ In the gnder oqofosof fth varius s In Kenya food was judged to be non tradable In that in a normal year the country would be selfuffident This implies either that women have differen- secws in which mwan wage rates are lwestf An Idear potyl of the t tially lower raw of accumulation of human capital, or that they face diffeentially high entry bar- Male of riers to the more remunerative sects (or both). Hoo" 8.18 A frther supplement is to disaggregate wage rate information by gender (where sample Fauwi Mde. Wap sizes permit). Controlling for hunan capital, _edw km" _ dal hb row differences in wage rates by gender seem to be Piseedd tadaed x y x Y x y x Y rather sma relative to the very large gender Umpmaced radable X Y x Y x Y X Y differences in labor allocation by sector. How-?Jr.dahla, ~ ~ ~ ~ ~, ~, ever, this remains uncertain and so it is worth Nonftda getting direct evidece upon It where this can be capiagob x y x Y x V x achiee at low cost. he two supplementary PW& sicrs x y x y x y x Y analyses ae mmarized in Box BA. Xtaiip2ed bsuhcategmy Box 8A and Frame 8.2 takeus quite a long YThwif.27w Wchf(tbfw way in detmiing whether Send& is likely to An aetual example for mmy, ibe Pa.d lo$rc. a significant issue in structural adjustme (from the viewpoint of production). For example, they might between them suggest. sowm'... * that most of the labor that must reallocate from nontradables is female; U _pwtradams * that this labor is not allocated thwough the Waelaboone sta market; P noteftdbwaum 12 es e that this labor appears to be locked into the Nowdal sector in that It already has a typically low earne4d 63 v ings; and that the problem (if any) in the labor ftuift mvke Oup mart appears to be one of differential acss Non-marba,bic uerks) 19 e1 rather than of differential earnings once in the market. &UVw ~ dtaamtav.2.t'w nit ~duiui Mala*uiat' 8.20 Ihe constuction of the tables should be Ch un r. ioqst( feasible on many sets of survey data. The only disaggregation which will soetm be problematic is the allocation of labor by gender beissues If the hierarchy of wage rates by sector tween food and export crop agriculre. In case tends to coespond to the concentration of fe- whete surveys do not recid labor allocation by male labor In each sector. For eample, we might crop, a proxy can still be achieved by investigatfind that women tend to be concentrated in those ing whether, for each of male and female haded 155

161 houshd, there is a correlation betwee the lady ti aspetofresource mbility from prportin of crop output which is export crops, the view point of gender is the p oess ently and the propoion of the houshl labooce Ito emnomic activities which the agent has not whihismale. Ausefulte quefor invesdgat- previously undetakn. This is likly td be of ing this sort of question is a regression In this Inport where women are heavily concencase the regression would be: trate i those activities which need to dedle relatively. In much of Afrks the central process y aa + of resoure allocation is the expanon of export agncultute. We first give an example whidh sugwhere y 1 is the proportion (by value) offarmout- gests that thee can be a substantial, gender-reput attributable to export crops in household I lated problem in this expasion. The example and x, is the proportion of the household i labor concerns tea-growing in Kenya. fore which is female 8.25 Recent work on the adoption of tea in tea- 821 Two things should be stressed about this growing parts of Kenya during has comregression. st ft is very 'reduced form." That pared male and female-beaded households. Tea is,itinvestigapteswhingwhchistheoutcome is arguably the most Important export activity of a more complex, but urnodeled, process. with potential for expansion in Kenya, since un- Activity choices can be presumed to reflect con- likecoffeeitisnotsubjecttointentionalquotas. slderaionsoffactorendowmertotherthanlabor, Hence, in investigating the deminants of tea suchaslandandcredit,andalsoconsiderationsof adoption we are at the heart of impediments to dsk Potentally, the regresion can be expanded appropriate resource mobility. The study found to il sude such influences. Second, much choice that el-headed households had only half the remains open to the analyst: for example, how to propensity of male-headed households to adopt definethelaborfore;whethwertousegrossoutput tea. Since in Kenya around a third of rmal or net income, households are femal-headed this diminished 822 Non-labor fatora of production are usu- propensity is in aggregate subsaal Further, ally mwre difficulto d teby gender from the case of Kenyan tea is particularly reveaing surveydata. Asanapproxmationtheagent who because most of the tea picking is done by fecontrols the allocation of land as between food males. This is reflected in the effects of the and export crops can be tak to be the head of houseiold labor endowment on the propensitq the household, although the exent to which this to adopt tea. Holding other household characis inacuate will need to be checd from non- teistics constant, extra male labor has no effect urvey soures. In this case the reevant gender upon the propensity to adopt tea whereas extra dilsagtion forland isbetween male and fe- female labor leads to a statistically signiflcant male-headed households asshown in Frame 83. increase An additional female adiut in an oth- 823 Control of capitdisaregid bysector erwise average household raises the propensity and gender is not usualy identfiable fom sur- to adopt by around a quarter. Thus, in Kenya the vey doaa key sector of tea is dcaacterized by three appar- 824 So far the analysis has concerned the static ently incompatible fcts. Women do most of the picture of resowre allocafion. We now advance work on tea, households with more women are to the analysis of resource mobility. A particu- more likely to adopt the crop, yet households headed by women are far less likely to do so. The implicaton is that female-headed houselds face some severe constaints additional to those faced Prm -3 The dlocatio of lad by sector bymeale dedhouseholdswhichprentsthem and gender om entering what would otherwise beanasuri M Fank4Ui" activity for female-headed households 8.26 How can the policy analyst study this sort of pheomenon? The most stightward tech- Foodap x x nique is tabulr and this is the technique which otcrepe X x we have relied upon so far. Box 85 shows how Oh&rWUMs X X tables might be used to investigate gender effects Xb 16 _iaampwmtw of lmd urd fwwepwpml in tea adoption. However, the difficulty with using tables for this sort of question is that sev-

162 Box 8.5 investigating gender influecem an tea adoption using tables A hypouledl aomple An eample of. tuburpprach tobu adaption 7ep*blm: ToinvesWtebytaebubrn whergen-.a.s i vsl) # der sa slgfiantflue onotheoadoption oftea The linitao of using tables is that tee likey r to be Mlabog _ d-wamet I Z at I* seva iuec on to adopton. Suppese we statwilth Pa_ elabore.dsowammiladalrteis the hypohesis that the age and gender of the housod Ma1eheaded X X X are lkely to autter as are the houseds endow_mat of Fimal-heade X X X made and femal labor. Tables an te be unsd to see Pm orea n2or3 whethe there esubstantia diffeeas in the prporto Mal-beaded X X X of houeolds adopting tea by breakg the sample down Fead labor ens ometeded 3 X Xe X into "ee within households have fte same ae group M1.0i"led X X X and gnde of th head, te same niale labor endowments ue -headed X X X re eah divided Ito thee goups. With 3 ag groups, 2 blaloremdowmetalogsadullts gendersm 3 male labor groups and 3 ftne labor groups X Xde_aerk X the are threfore 3O3 = 54 cells. We tien ampue x x the po i of h _oueholds adoptng tea in these as. emlela2orx The tabulw apach rapidly bits the limwis of sanple size leheaed X X X if there are sever ifluensa to be investgted sin the Fenmae.eaded X X X preset example. The four luences a only be handed Feranelabor endaw_rtt niothman3 bogew by breaing the sample down Into 54 cell, but Male4maded X X X am tee must be enough obsenvaio in each cell for a Femalbe-eaded X X X proportin of adopes to be a meaninsg figure. F Pe abelead.w.er_ajultaor lerl samples in the 1.0M to 2,00Q ramge brea the sample Mae-bed X X X down nuch beyond 50 cells i unlkely to yied reiable Fmale4heded X X X nfmatoin Hence, in the above exmnple, If there were Female laborotdowment2ar3 more than four Important Iluences on tea adopion It Maeded X X X wouldbedifficulttoinvestgatethe effeds of genderuing Femal-hea_Ded X X X Uig Female labor endowmentmonet1an3 only tables The alternative apprach in such a situahtonis _hed x x x discused in Box && nhd x x x 190S.sdoged ftuda5.r el difernt influences need to be disentangled. rather than a regressi, although a mgression Tea adoption is, by hypothsis, influenced by the can be used if no software is available to run a gender of the household head, by the male and logit Regeon software is univesal and the fmalelaborendownents,andprobablybyother results are straightforward to intpe Logit factors such as the age of te head. Each te an resuts are more difficult to intepret. Box 8.6 influence is added requies a fuher breakdown (overleaf) sets out, using Kenyan data, a logit of the sample and this rapidly ehausts the sample analysis of tea adoption. size becuse the number of observadons in- 827 Logit analysis is appropriate whev the ing in each cell becme too few to be a relable depeden variable can only take two or thuee guide. Hence, economists tnd to use more pow- values. However, some processes are a mbtue erful tehnkiqes than tables when confronted by of this and a continuous regressio An impormore than a few potental influncesr Two hghy t4nt example is erngs in the laborn.rket Such useful te ques are rgessin and logits. Re- eaings reflect two inter-related process gaingresonsareappropriatewheret hepoeno ing acoess to a wage job and getting a partcular to be exphined can potentally take on any value wage rate in that job. To analyse gender isse in (such as income). Logits are appropriate where the labor market these two processes are best the phemon to be explahwd can only take investigated simultaneously. The apprpat on twoor thee values. Fbrexample,in theprest methodology is a tobit, which combines the discase there can only be two possible outdmes: creteoutcomesmodeledbyalogitwiththemore either the household does or does not adopt tea common continuous dependent variable rees- The appropriate tecdhque is therefo a logit sion analysis. For an example of such a gender- 157

163 Box U6: Gender effecds on te. adoptim: The loglt approach Wh Othee are several tnruenow of te adaptn the d whi thi probability gos toug in a tabla approa-h enountes its imit and moreporful oitrt, te probablity of adoption con -may vawy be tednque ar needed Whe the pheomenon to be e- twoe Oand I but for stattcpalpouesi t is bettri the pdlnad an tk many valem the nom iodhnlque to ieg8lul Tea adopo, howev, is not lme tdhi we are "dependent varable tobe able to vay bebtwem- Irniy ad. Iqf,nty. A lot adiev this by expaiwog t te investg a only rig two stato, adoption onr pobabilty, P. but PA(-). Fr t averag ousehold In TheWdmquevwlch dealswithuch problemistemed. our eyn ample tis was 04/0 53= Q89. Secard, a gi*t It available on many stastical oft-ware packa. lgitworks with logaithb, sofor te average_d Here we explah how to use it and how to tpret results We msate wth Kenya data e the dependent variabl is the logrithm of 0.89, WM s Now we arereadyto Interprethe nder MeIe MM d Rual Kenya 1982 From the ress column we see that this h ltbs tells The hypo_s Tmea adopto is inhaened by the ae and gender of the hous-&old head, the male and fmale labor us tat6 ora househo with otherwise aveagehaac- t If It Is fmale-haded Instead of bon male-headed, endowmnts of leehouseho thepoporti of neighb- nlp(l-p)] insead of bing wm be reduced by 10, aghouseholds dopttea tnd wheth theho d Is tatis itwmibe Wenowworkbad to the t alreadygrowfg coffe The mulft. of adoptm which thi impi h, we move 6im the logaithum P/(I-? baek to P/I-P). The antilog of is V-wkbk C*4dmd r-tqu* 031. Se4dwenovebafiomP/(1-?)toP. Sacs P/l- P) is 031, P is This tells no that otherwise averge Houmlioldhaa (eadher,.s 2as h oolds, eutae hadedness lowas the lield of adoption of team from 47% to 24%, an enormous reduc- Age AgeoaWum Am05 OA04 IV to6 ti The effect of adding one more female adult to the ava houeod is to alseb4p/(-p)l by0.o42oitrises from 412 to +03Q Theantlog of 0Ois 1JS; so thp/a- LAboab lawhunt F) = 1.35, P =.S7. The extra feale adult thus inems MeANs the l1kelihood of adoption of team frno 47% to 57%. Bath Pezwhs a of the effeeb are st cahlluy s1gificant as cm be seme N elhimfthous"do 42D 455 from the t-ratlas. By ontrast, the endowment of male A adywmwlng coffee Z0S labor has no effect upon adopon (the t-ratlo is insigficant). hptg Le s: The I sdents show the effect of The lognt approada is a much more sophisticated proce- 8 gawe kon t the lkelfood hat the household adopts tea dure tbular n analysis. tn betesftware and in tal 47% of the same adopted te during thepeliod so vare thgt but it i lusted when tbere are too any keheavreqaehousdthe probabityis Q47. Inorder Imptat influenes to Ihrporate ito tabr breiakto interet the cxffidentit is nsay to allow for the downs. focused analys using SDA data for the COte men to credit we would then fully accunt for divoim see Applet on, Colie nd Horall (190. theoservedie nbeavior. Inplactice, That analysis found that once in the labor market however, our lnasures of diffeential constws wom receivedthesameswasas nn,ontrol- ame unlky to be sufficient flly to acwout for ling for their educationd haracteist, but ftat the effect of gender. Hence, what this stage of the e among thse already in the labor force, analysis can accomph is primarily t estabsh woman were radicaly less likly fan men to whethe the symptoms of gendert-difentkatd enterwageemployment,and thattheirparticpa- consaints are prent and ninpornt. If they tlon was far more nsitive to wage rates are, there is at least prima fde a case for Ivesti- 828 In some respects, the ideal for a gender- gationoflkelyconstaintsandhowpoliesbear focued anaysisis tht as other elanatoy var- on them This is taken up in fte next secto abies are added, gender gadually ceases to be a sigrkifcant o. The hypoteusisis not Pour gendediffren processe that women and men have intrinsically different ewnomi atitudes but rather that they tend to 8.29 Four distinct posses account for why fiae dlflneetconsralnts Hence, we wereableto woman face differential constints upon ecogt satiatory empiical measure of these dif- nondc activity. Frst, womemay enownter disfernes, for example, better acess on the part of cruimnaton outside the household. In developed 158

164 countries the most emphasized example has which is cntrolled by men This gives rise to a probably been diffeental wage rates. In devel- cassic "principal-ag problem the oiman has oping comtries, however, discrmition in the littleincentivetoworlcwell. Forexample,Ongaro lbor market appers more to take the form of (1988) used a recent Kenyan sample swvey to differential acmess to wage employment For ex- compare the effectiveness of weeding (a female ampe, in rural Tanzaia (where fonral sector obugation) on maize yields in male and fewleemployment is an elite occupation) men with headed households. He spedfied a regression of secondry education had a 3-in chance of such maie yields per heta in whih the number of a job whereas women with the same education weedings was one of the explanatody variables had ondy half that chance& Often a more impor- and then fitted the regression separately for male tant insan of discrimination is the credit mar- and female-headed households. In both types of ket. Because usually women do not own market- household there were two weedns per season ablelandrightsandassubordinatesinthehouse- and each we. ing signficantly raised yields. hold cannot estabish reputations for credit-wor- However, whereas in female-headed households thiness, they tend to have mrkedly worse access these weedings raised yields by 56 per cent, in to credit. This is especally severe in economies male-haded househokls the increaseinyield was sbect to financial repression where credit mar- only 15 per cent. Since other differences were kets are rationed by risk-bearing ability, women controlled for, the most likely explanation is a are pariularly disadvantaged, systematic difference in effort due to differental 830 The second process, whch also operates incentives. Thus the"inntives" argument, now outside the household, is the different directions so familiar in the World Bank, does not stop at in which the tendency to imitate or copy role the door of the household. To put this in pernodlds att*ats men and women. The tendency spective, if Ongaros sample is representative of to let dedsions be influerced by what other simi- rural Kenya, the national maize loss from this lar or adnmed agents have chosen is a universal disincentive effect is about equal to the naize feature of human beamvior. It is a key way in gain from the application of phosphate and niwhich irmovations spread over the population. togen fertilizers. There is now some evidence that role models are 8.32 The fourth process is the burden of reprogender-specific: girls odpy women, boys copy duction. Because there is a phase during mid-life 3m For example, in urban Cote dlivoire people in whch womens time is pre-cnuitted, certain have been found much more likely to enter for- activities are preduded. Skills decumulate, and mal wage employment if their parents had been long term contracts such as are conunon in the so occupied (controling for other characteristics), labor market are teminated The physical debut there is some indication that females were mands of child bearing and breast feeding stain not itrluenced by their fathers occupation and health recent studies (Bevan et al, 1989) show boys not by their mothems each tended to copy that female health relative to male health goes the parent of their own gender. An implication through a trough in the child rearing yeas. This of this is that if some new economic opportunity health deterioration rebounds upon income is initially taken up by men it may automatically eaming opportnities, especially due to the unbe diffused over the male population by a ceraindiscontnuities in the availability of labor. nmehanism whidh wil not tranit it to the fe- Wormn become confined to a range of economic malepopulation(appleton,collierandhorsnell, activities in which such discontinuities are rela- 1990). tively unimportant. 831 The third process is that within the house- 833 Between them, these four differes from hold there are asymmetric rights and obligations. men tend to skew female labor arocation to dif- For example, in ral Africa womenincur obliga- ferent sectors from men, and to impair female ions to grow food crops for subsstence, to gather mobility between sectors, the symptoms sugfuel and water, to cook, and to rear children. In gested in pargraphs 8.09 through 828. return, men meet crtain cash needs of the household and usuay are responsible for the alloca- How public polkces relate tion of land. This pattern of reciprocal obligatons is often unequal. In runl Africa women 834 There is a case for skewing the provision wok for considerably longer hours than men. of certain public services towards women, be- Part of this work is on holdings the output of cause private substitutes tend to be skewed 159

165 against thm However, curfently public provi- ual extesion visits, but most farers have the sion itself often tends to be skewed against chance to attend group de atons (whih women. In several key processes private maket involve signiicant time and traspot costs), and aid non-market access mehanisms tend to favor to copy their neighbors The determinats of men over women and these differeucs are rein- access to each inmation channel can be invesforced by public allocaion Wome of dcdld- Opgated indeendenty MOd, more Impotnty, bearing age have amarkedly higher incidence of their (ntection can be studied. The public iniliness than men, yet a markedly lower propen- formation chanel, extension, needs to be invessity to use public health facilities. In the work tigated in conjunction with the private inornmaon the adoption of new activities, Bevan, Coblier lion channel, the copying of other households, and Gunning (1989) found a powerful local copy- which is probably more powedrul The SDA suring effect households imitate what others are veys can provide both subjective and objecdve doing. If each gender tends to follow its own role information on leaning from the exteion servmodels, women have fewer opporunities to leam ice and from neighbors Subjective infmation through imitation since the new activities are canbegatheredontheresondentsstdsoums pimnarily the preserve of mel Yet the public of the information on which canges in agriculextension service is usually directed mainly to- tural techniques were based. Objective data are wards men. The propensity of paets to pay for available on contacts with te extension agents the private secondary education of daughters It is also possible to use the duster-feature of the appears td be substantially less than that for sons. samples to investigate the effects of neighbors, a Yet the maprityof pubic scondary school places cluster being a contiguous group. rence, it is are allocated to boys. We will explain below how possible to use information on other households to quantify the effect upon the private demand in the cluster to investigate copying of neighbors. for secondary school places of a marginal reallo- 837 It is then possible to analyse whether the cation of public places from boys to gis. Turn- are gender biases in the pro_ss of infrmation ing from the distribution of goods and services to acquisition. If there are, the biases in private and inputs, women work markdly longer hours than in publc processes can be compared. men and yet bear the primary obigation to provide cmmon-use inputs to the household (water Health semices and fuel). This obligation can be exremely time conmrdng: whether it is, depends heavily upon 8.38 The location of households relative to public inrstuctur polcies. Women, lacking health dinics varies so that some households will autonomy and land rights, face greater difficulty lace prohibitive costs of access Public and priin building creditworthiness reputations and of- vate clinics also coexdst, with differens in user feting collateral. Hence, tey are at a disadvan- charges, lime costs and quality. With individual tage in private credit markets. Yet public credit dataonsymptoms,actins(includingtreatmet), programs tend to be orientated towards men. user costs and the duration of illness, it is po- 835 The genenalization from the above ex- sible to estimate gender-specific utilization of amples is that a key distcion as to service pro- health see functdons. It is also possible to vision is between those services which are in investigate the determinants of child-aess to rationed supply (generaly public) and those health care, in particular the effects of poliywhich are cleared by the markel Secondary depedent mnte carhe Immeeducation, information about techniques, health diate policy quesonsare: care, water wpplies, and credit are all partly ra- (a) Are women less likely than men to make use tioned and partly provided privately. The SDA of health serices fora common level of scknes? data can be used to Identify the gender-pe- (b) If so, do policy vaiables such as disne td cific determinants of acess to services, both te facility and the level of use charge have differdirect deteminants of rationed and non-rationed ential gender effects whih would enable this access, and theintracions between rationed and bias against women to be offset? unratoned provisi (c) Matare the intacons betwe health care prvison and other policies? For exmple, is Extenimsero esi ad cying of techniques child use of (girls/boys distinguished) health services increased more by a greater geographic 8.36 Only some famters have access to individ- density of hulth cinis or by a wider spread of 160

166 femaducation? Piwp twa Birth conbul SAl Somesurshavedetaledldnormaon upon wet and dry season watersupple Some 8.39 The burden of reproductn and dhild- households have acoess to cmunay piped rearing isusualy measured in termsof the num- water, otes to a private tap, and othes only ber of hildren in the househol Howeve, It to tradftiond sources. Piped water can have nay also be a contngent ability. Households in powef fows r both female labor which women of childaring age do not bave supply and famiy healdt Henoe, Its distrbuaess to rliable methods of birth control face a tion Is of some Impotane While accss to probability distributon of the numbers and tm- private piped water can be viewed as exogitg of fture children. A potentially Importat enous, acess to mm piped water is to contribution of birth cotrol is that, by making some extet determined by the household, althese rsks controllable, the household is feed though households will differ as to the tie from the need to guard against unorain futue costs of access. The rlative effects upon use liablities. Hence, birlt antrol, by reducing rsk, and time saved of dierig piped water stratecan altercurrent investnent sttegim Its effects gies constitute an Irtan policy Issue (use Is can thereom be nkediate ratherhmnbeingordy mwasured in terms of both of ftme spent collong tem For example, one influence on Invest- lecting water and the quantity of water conment whkh has recnly been invesgated using sumed). African survey data has been the effct of future birth lbties on the decision of the household Credit to Invest in the private secondary education of Its curret hldr It was found that, controuing 8A2 The SDA surves include some inforfor the number of caunt children, the age of the manton on credit. Asming that ths informahusband and household ncme, the future birth tion is at the household rather than the indiliabilities of the household powerfully reduced vidual level, gender issues can only be investithe willngs to invest in the educaion of chi- gated through a omnpris of femade and male. dren. Similar powerful effects may well influ- headed households. The information can be ence other household _ such as In- used to analyse the determinats of to proved livestock, tree crops or non-farm enter- both private and public credit (and teir nerprises These can be investgted along the same action). Poliy issues include: lines. (a) Do female-headed households have differ Teexpeted future bh liabtes of the entlay poor access to the private credit marhousehold ae not directly observed in surveys ket? but can be readly proxed by the ag of the wife. (b) Do they have differentilly poor access to It is possible to investate this effect by using public credit? tables. Using the same tehnque as in Box 8s, (c) What chara scsenablef aded househl can be grouped ito those in whkh households to gain access to public (private) the wife is still of cild-beaing age and those in crdit? Are they the sae, or Is publie credit which the wife is above that age (n households being targeted to those shut out of the private with multiple wives it is the age of th youngest credit maket? wwch matters). These groups can en be furher divided as in Box 85 by the age group of the Eduction household head, labor endowments and whatever else Is considered to be important and fea National educational emol t figures sible given the samnple size. Having split the are usually avaiable by gender, and thes prosample into groups wich ate comparable in ve- vide a srng point for the Investlgation of spectsotherthandtild4beaingtheffectof cild- genderdiffered However, theycanbequite bearing can be discred from the different pro- mlei and household survey data can be pentles to invest. Again, moe poweful tec- parcularly useful. Take access to primay niques than tables can make the analysls more schooling. Data on schiol enmt will rlable and take It further. typically show that boys outnumber Mrs but not by a wide margin. For exanple, in C6te 161

167 r -ox 8-7: rimary -o d'ivoire about 44% of enrollments are girs, 56% Box M: Gender and accss to pdmwy boys. However, since most chikren go to sdhool schoot sefaily smal dffmwes in ollment figures 17ws iasis SDA survey of COt. divolre hide lange gender dffrences in non-enrolment w pmbklw To inves ptte V t "uences X Among those who never go to schwol girls make am prhna ysciholing forb andgi up 66% and boys 34%. A 12% differee in en- Ik tdudq: A logit analyis (see Box 8.6) was e. roilnent rates hides a 32% difference in non- Bele it was important to Ientify differene in how enrollment rates. aearly, the policy m anker onincome affected the ar of boys and girls, the sample cerned with extending educational aoess needs ofchirn wassplttntoasampleofboysand sampleof to focus upon those children not gon to school, gir and the logit estimated separately for ea sample. and most of these are girls. The policy maker The rest awe shown in t able below. Althoughe next needs to ask why these dhildren are not tabe ontains valuable information, b isa very poor way tof plesnfia : thecoe iden sq ttdhedtly going to school A rly in t queion be tpretd and there is too muh information for the is whether it is mainly a matter of low income, in poic maker to sm what is Important. Figure 82, how- which case the problem is part of the nuch wider ever, represents part of the infomation a much more one of poverty, or whether it is more specific. digestible way. h shows how the proabity of non- Since there are many influences upon whether a enwolment is related to hou_od incme for each gen- child is enwoled, the tabular approach is not likely der. Iti _cuatd for the otherwise average d. That to prove very useful. As with tea adoption, dnce Is, we set all other variables such as th aev Of the householdhead,equal to theaveragefor thesample, and the event we are analyzing is "disret (either vary only houehold Ineme The coeffidents on house- the child does, or does not, go to school) the holdincomeareusednthewaydesribedinboxb86)to appropriate technique is a logit. Box 8.7 hows cuaos how the pro lty changs As incomes rise, how data for the Cote d'ivoire were analyzed boys become abnost ertain to be sent to school. Gls using a logit and the results turned into a simple from very low4ncome households have a 33% dhance of chart. The key result shown by the chart was that to about 22%i but beyond that Itcoes fl 'o fu if incomes rise sufficiently virtually al boys go to (adhto a t2heyda o tat ay dses ainthoughuthe school but around a quarter of girls remain researchr a' probabydismdss tisasreflectingthefew uneducated. The problemof uneducated children observaton among very rich households). This teus us ends up being entirely one of gender as developthat bm aloe was not going to solve the Ivorian ment proceeds. Since income was not going to probkm of uneducat chfidren. solve the Ivorlan problem of uneducated children, lag estimates of deteminants of en o in the government was probably right to make priprimaq school age 11 yeagbrsy m mary education compulsory (a decision taken _to_dy G"wdy after this survey): otherwise many girls would Girls o#ul, grow up to be uneducated mothers which we Y i*k G< r-ciaejicina dm T-nU. now know has deleterious effects for their children. Canmast 4Lt Survey data can also be used to investis_d 43M e -ZWS6 gate the whole educational hierarchy. For ex- Nolvrlan -L laos ample, in the CMte d'lvoire, there are also large North differences In the chances of boys and grls who Urban 0G? z Mottereducatd complete primary sihooling advancing to gdv- Father educated einment secondary school. Places are rationed Gv tb oai 263 by an examination. Appleton, Collier and hpmeperjob a e el 2339 Horsnell (1990) split the problem of access Into Squad OM16e e4B two stages. performnuce in the examination and Cubdw 4L230-0.o.M341626e40S 1.7 being sent to school conditional upon examina- Lvestodcpwcaplia Se ti on performance. They found that the lower 9owwAppM*aS,P.Co1 andp.ht_ca DWENUM ia chances of girls were entirely explained by their worms performance in the examinatdon. Know- ing this, the policy maker would then ned to find out why girls were leaming less In pimary U IA nmvat saditawsam NfPas1p#to AsAnayI a1tmitefic;timsd:a.wwwi B9mIIW Table&Si. 162 school than boys. The answer might be related with the school teaching system or it mright be a

168 Ptababg of / 02= D 75 lcd Proabifties e'aouatd at the mean of olw regremsww Rat"of piobabgites U10 o nsbdmffcscvnj~ald at the meam of oue vwemw &$m Apttm 4 ho CBtsole P. CoOwWd P. Ho tne: Gender DOUres in Edutad=4 Attuinmeat m1d in La Marut Paadpatm As AiWyas of dv,sda WaM Bank i. FIRUro. problem at home. Recafl from Box 83 h-at In Kenya school-age girls were spendilg far more dtme hanbos contributing to household Incomc. 1. SwQdlIe Rbdwan antd Wwngwe (96) t03

169 ~~9 Smaliholders and structural adjustment 1rdroductlon and meso-economichanges on smalwholders as prducrs, labors and consume he aalysis 9.01 Agricultural performance in the gret of food CDnsumption and nutrition, and access to majority of countries in sub-saran Afica has, health caue and education offered in other,chapin rent years, been a source of conen both to ters in this Volume appli to the case of smallgovernments and intemational Institutions. The holders and need not be recapitlated here. The combination of indement climte, fragile ecol- sameistruefor thepovertyproflleandtheanaly- Ogy, utable world markets plus, in some cases, sis developed in Chapter Four on wage employthenegativethnustofpastpolides,hasproduced, ment and ral non-fm household enterprises. with some notableewepoms, thedisturbingsce- Howeve, smallholders are typially both pronaro of diminishing expots and escalating im- ducers and laborers, and dthy produce partly for ports of agricultuanl commodites In per capita sale and partly for subsistence consmption. terms there has been an overal decline in pro Agricultural production activities are not duction, and, in some Instances, the proportion studied in any of the Analysis Plans reported so Of output marketed has fallen. far, nor Is the interdependence of decisons to 9.02 Theoverwhdmingmajorityofagriculural allocate labor to the famdly plot as opposed to offproduers, and indeed the major part of the rural farn employment. (The simultaneous detenniand tota population, can be dasd as small- nation of subsistenceqcnsumption and purchases holders, ie. peasants whose main resource is their of food was mntoned in Chapter Seven). This own labor snd whose plot sizes are in many Chapter, therefore, focuses on (i) the relative African settings determined by family labor con- contribution of agricultural (or crop and livestraints rather than access to land. Thus, any stdck) production activities and wage labor to changes in the ecnomc and social welfare of household Income and (0) alterations in the allothis sector have a profound impact on the long- cation of time to on-fam and off-farn employterm petformance of national economies A spe- munt owing to labor, product and input nmrket cial interest In smiholdes justified because changes associated with adjustment. Diter they constitute the bulk of the poor in Africa and respones in terms of the household's allocation becuse of the potential sigificance of their con- of its labor resources naturally lead to differences tribution to succesful adjustment through a sup- in consumpdon and welfare indicators. ply esponse As discussed in Chapter One, It is impor From a social dimensions point of view tant to recognize that the SDA Integated Survey the broad anaytical issues to be addressed in the isnot a dedicated agricultural survey designed to smalholder context are uchanged lative to obtain detailed, year-round farm management other groups. They Include the effeds of nmcro- data and crop-speific input-output (hinduding 165

170 labor use) infornation. It is therefore not an 906 Structural adjustment, as described in this appropriate surce to study the agricultural sup- Volume, aims to depredate the eal exhnge ply response under adstment It does, how- rate by reducng import restrictions, nominal ever, permit analysis of labor allocation between devaluation and reducing inflation. Devaluation off-farn and aggregate on-farm employment, and exdhange unification are expected to help farm profit, the use of purchased inputs (an im- reduce the tax implicit in the exchange rate reportantpredictorof agricltural productivity)and gime when farmers are forced to sell at an offichanges in assets and the subsistence ratio in cial, overvalued rate, while buyig consumer response to appropriate household charactis- goods and inputs at prices rflecting dearer partics and meso-economic changes. Thus, the allel market rates. Trade liberalization can be Chapter is structured as follows. Paragraphs 9.05 expected to lead to an incease in the availability through 9.16 summariz key mac-meso link- of aonsumer or 'incentive' goods, and thus a ages as they pertain to smalbolders. Paragraphs higher supply response (Bevan et al 1989 show 9.17 thuough 9.68, on meso-xicro analysis con- that rationing of conumer goods sevrely conslstsof sub-sections onth classification of small- strains peasants' ability and willingness to proholders, the analysis of access to productive re- duce a marketable surpls). Adjustment may, sources, the analysis of smallholder terms of trade therefore, be considered a necessary condition to and coefficients of net protection, the demand for reverse the process of agricultural decline referrd agricultural inputs, the analysis of farm profits, to in the previous section and to improve the and the simultaneous determination of on-farm income opportunities of snallholders, as long as and off-farm labor supply. A brief conclusion is cost increases do not outweigh producer price offered in paragraphs 9.69 through increases and an adequate share of devaluaion- Throughout, smallholders are treated in gender- related price gains are passed on to farm Proundifferentiated fashion, since approaches to ducers are, of course, negatively affected by degender differentiation were the subject of Chap- valuation-induced increases in the cost of imter Eight. ported inputs. But against this should be set the costof overvaluation,includinglong runnwesource Macro-meso linkages' misallocation and shortages of consumer goods and inputs resulting from chronic foreign ex There is wide agreement nowadays that change shortages. Because of the low import the causes of Africa's agricultural difficulties in- content of their input mix, smallholders would dude inadequate price and institutional policies seem to benefit disproportionately from devaluof the kind which structural adjustment programs ation relative to larger farmersm On the other seek to correct. Agriculture is the main sector hand, risk-averse producers may perceive inproducing tradables in most African economies, creases in the prices of purchased inputs as inin the form of export crops and food commodi- creasing the risk of inmovations capable of enties partly for export and partly for import substi- hancing productivity. This would, however, aptution. "Urban-biased" development policies pear to be the case only where nominal devaluaiming to promote industrial import-substitution ation does not translate into real devaluation, for through trade restrictions tend to discriminate example, owing to insufficient monetary and fisagainst agriculture by "closing" the national cal restraint. economy and raising the price of nontradables 9.07 Thisbeingsaid,itisimportanttopoiatout relative to that of tradables (an appreciation of that adjustment in terms of the variables which the real exchange rate). Indirect agricultural affect the real exchange rate is far from suffcient taxation associated with overvalued exchange for agricultural recovery and rural povety allerates in Africa has often been accompanied by viation. It must be accompanied by public inhigh levels of foreign trade taxation in the form vestment in agriculture and the development of of export levies and monopsonistic, state.con- institutions capable of serving the needs of rual trolled food procuement.thedisincentiveeffects people in the areas of input distribution, credit, associated with these price policies have been product marketing and social services-all tese reinforced by the low priority accorded agricul- serving to provide an "enabling environbme." ture in public investment policies, particularly in These measures take time to implemt A disthe areas of transport, apprwiately-scaled irri- cussion of the short-run prospects of the rural gation, and agricultral resewta and extesionl poor under adjustment is, therefow, in order. It t66

171 is appropriately developed with reference to ru In zemte areas, food crops nuy be nonra labor markets, the dege of tradability of tradable due to a lack of transport ne food crops, access to productive assets and fra- and furthr exacerbation due to lack of repair structure, market Imperfections distorting the and mate of the existing stock following price signals of adjustment, and the factors which short-term fiscal retrenc nt. The chronic poor hibit the adoption of high-return activiies. A in these areas who produce nontaded foods wil word of caution is howeve necessary from the seeno inrovementintheirprlcesresultingfrom outset The discussions below present the more devaluation, but no decline either. As investgeneral senaio and it is very likely that there ments under adjustment improve the mesocould be anomalous cases which need to be iden- economy in the medium to long term gradually, tified and supported by countervailing policy to emote regions will be drawn into national marbridge households over the unintended and un- kets. LmoI prices will inaeasingly be affected by expected mnsequences of policy reform national (and world) market conditions As a 9.08 As wage eamers, smaliholders are likely result, nontradable foods wfll become more to gain in the medium to longer tenn as the de- tradable in dhwacter, and the chrdc poor should mand fortheirlabor(inthecultivationof tradable benefit frum devaluation as producers, if they crops) ises. Rural labor may well have a low can market some surplus. supply elasticity in many African countries be Thisassumes.of ourse, thatthepoorhave cause of the relative absence of landless laborers. access to land and security of tenure remains in The seasonaty of agriculture forces the wage up tact. Securing the claims to ownership that poor at the busiest times when everyone has a Lbor people have over their assets Isan important first shortage. The real wage effect for nrual wage- step to ensuring that poorer rural groups benefit earners of devaluation would thus tend to be from an adjustwent program (Chambers, 1988:3). beneficial over the longer term. This is especially Land rights are often tenuous in customary artrue where the adjustment program induces sus- eas, and the poor frequently have few chaels taned agricultural growth. for redressing their grievances. Unless policy 9.09 In the short term, however, rural wage reformisaccompaniedbymeanresensuringthe labor might suffer a real-wage fall. This would right to land by poor households, large estates occur if estate production is slow to exand and/ maindy growing tradables and benefiting fom or if nominal wages rise less, in the short run, policy reform start a spree of encroachment and than the cost of key wage goods. This is an land hoarding for speculative purposes. This empirical matter. There is anecdotal evidence will lead to a gradual diminution of land ctltialso supporting short-run real wage deteriora- vated by the smalholder sector. Thus, extendtion mediated in a slightly different way. Rela- ing property rights to smaliolder households, in tivelyrapidestateresponsetoincentiveshasbeen addition to ensuring their subsistence, will raise accompanied by increased demand for labor and their incentives to maintain and improve those higher nonual wages, but immediately offset by asse. When land tenure interventions are made, escalations in the cost of living. By and large, it is care mustbe taken to protect the traditional rigbhts uncertain wheter smaliholders would require a of women to cultivate land for food. Such measreal-wage fall to induce their expansion of tradable ures in the past have often designated men in the production (this is complicated by the fact that household as title holders, thus weakening the they supply part of their wage labor to each other). claims of women. C6te dlvoire, Ethiopia, Kenya The short-term effects of devaluation on real and Zimbabwe have now given women the right wages are therefore ambiguous. Some food to inherit and own property (World Bank, intervention to the rural poor or other compen- 1986d;40). 2 satory assistance may be needed, espedally on 9.12 Improving the access of poor groups to estates, and to the net-bying smamlholders and Infrasuchtre can be as benefical in many cases to rurl labor. The scale of this will be less than as increasing their productive assets. Infrasructhat which iscalled forin urban areasand donors ture investments often bypass areas containing such as the World Food Program have extnsive high concentrations of poor people. This raises experience in food assistance to estate workers their production costs, and acts as a barrier to Tlheirprogamsmaytheeforeprovldeappropri- gains from greater specialization. For example ate avenues for food assistance to smalmholders while adjustment programs in Ghana and Kenya and/or rural labors under adjusment have sent out clear signals for the expansion of 167

172 tradable activities, many of the poorest farms whch nuyrequire m ypolces, such hve fad difficulties in achieig ths because as targeted extesion sevies encouraging fartof their lotion (Helleet al, 198&20). In making en to change crop mixes It b possible that some new infuc investents, policy-makess fail to benefit from adsustmentsneed to give more weight to actions wdch assist ply because they do not switch production tothe poor and open access to them In many areas wardsmhigh-retunacdvitieslnstaneatouy. Tids transportinfrastructurehas broken down, so that may arise initially fm a perepon lag- farmsome loca markets are powrly integrated with ers may take time to observe relative price the natonal maket, leading to lre regional price changes, and to expect them to continue. Or the difeen (Abm ed and Rustagi, 1987:109) whdch failure to change output patterns may be more is further compounded by administrative restric- deep-seated, arising from either ecological contlon on the flow of products. In designing a straints or risk aversion. The partcipation of protam of n r, opening access and poorhouseholdsinthemostprofitablecash crops rehabilitation with greater priority for areas ill- is usually below average. Such participation is a seved by transport infrastructure, but with a strong detemrdnant of income differences across high concentration of poor people, may be called households in most counries. In Tanzania, for for. example, the income of the poorest 50% of the 9.13 Another set of issues is raised when the village studies by Collier et al (198675) were failure of poor groups to benefit from adjustnent dependent on msubst crops for 70% of their wises from weak prce signals. In cases when income - this is double the share for the betterthisisduetoweakmarketingstructures,theterms off half. In M6te divoire appodxmately 44% of of trade facing poor farms may be raised the rural poor cultivate cocoa or coffee, comthrough improving the efficiency of offidial mar- pared to 65% of the rural population as a whole keting organizations. In Asia, for example, 75- (Glewwe and de Tray, 1987: 20). In Kenya the 90% of the cosmer price of food grains is paid probability of a household being poor falls if it to the farmer, while In Africa the proportion is grows tea and coffee - the main export crops ondy 35-60% (Ahrmed and Rustagi, 1987: 115). (Greer and Thorbecke, 1986). Although otton is Nearly30% of thediffertceinmarginsisdueto mainly grown by poor households in Cote the lower effiiency of African marketing orgari- d'ivoire, their participation in other, more profitzations. Inmanycountriesthegrowthof market- able crops, is lower (Glewwe and de Tray, 1987: ing costs has been as Important as cuevncy over- 14). Overall, female-headed households are less valuation in causing low producer prices (see likly to cultivate cash crops because the alloca- Harvey, 198&8 221 on Tanzania and Zambia, for tion of the nessary land, credit and inputs does example). Thrugh the integation of smalhold- not favor women. Altering the product-mix of ers into the mainstem, marketing reforms may female target groups will necessitate conscious now begin to benefit poor fairmer In Mali, for effort and inteventions in the supply of factors instce, a restrucuing of the marketing system of production to thenm has shifted output and input prices in favor of 9.15 In summary, raising cash-cropping by fiamers CuInenburg, 1987: 503). Imnprovements farmers offers an important way in which their in the effiency of marketing orgmaizati can bfts fom adjustment and growth can be inreduce the conflict of interest that exists between ceased. A number of recent studies have cdnpoor food producers and poor consumers over firmed that export cropping can lead to signififood prices. One study found that, for a sample cant gains (offsettng any unfavorable impacts) of African countries, reducing the marketing in smallholder income and nutrition (von Braun, margin by 25% would result in a 49% increase in 1989; von Braun et al, 1989). Kennedy and Cogill farm prices and a 13% fall in food prices, given (1989) have shown tht agricultural commercialireasonable assmption about demand and sup- zation in Kenya has had a favorable impact on ply eulicities (Abmed and Rustagi, 1987). landless agricultual workers, and has led to im If smallholders produce nontradables, or proved fainly nutrition and reduced hours cf if thy make intensive use of imported interme- work for women owing to reductions in the hod diate inputs such as ferdlizers (as stated above, crop acreage and increases in family incoe. In an unlikely event), they wll be adversly affected some cases the investmes in infrastructure and by devaluati The key to raising their incomes the inprovements in marketing discussed above lies in creatng altemative production patterns, may be adequate enoug for taget groups to 168

173 ralsethelrinomessufficlentdy. Inuchinsutnes, tion of labor (onarm or offfarm) are obvious poliy intwventons nust early be directed at definin and cassifcation charatefisib Both the mico level In other cases a more compre- affct the level of icme, while the cdmposltion hensive package of measues involving market- of hnome is influend by on-farm/off-farm laing and inrastructurd servies may be needed. bor use dedsions and crop choice (badable or Higher cash-copping may be dependent on ac- nontadable). The Identification of feasible and cess to productive assets being Increased in ft effectivepolcyiterventnstoralsesm klhdr ways disussed previously and on the extent to income requiesan understandingof thecompowhidch the market infrastructure is developed to siton of teir incone. Under citance of enable smauholders change their perception. relvely Inflexible factor proportions and un- Spedalization is a slow process but can take a hindered access to land, 3 the quantity of labor strong footing in the medium to long term available to the agrkiultual household will be a 9.16 In the foregoing, the most likely scenario sigdnficant determinant of farm inome. In many to emerge from polcy reorm and complemen- cases, therefore, the possblty of, and the necestuy policy measures required to sfeguard the sity for, hirng outside labor exists, and conserual poor has been dentified. But, the extent to quently the opportundty to augment housod which these measures are likely to nmterialize is income through Wring or selling labor. A relecontingent upon political economiconsiderations vant criterion in dffeentiaing betwe groups which have not been dealt with above. Where of farmems, thfore, is the proportion of housethe political elite have a vested interest in the hold to total labor used in production. In most growth of the tradable producing and estate sec- cases the employment of outside labor will intor, and less so in the nontradable food sector, volvethepaymentofwages,incashorinlknd,in and concentration may web emerge. The changes advance of harvest. But the poorer smawholders requred in the distrbution of fiscal resources wfll neither have the reseves from which to pay and the employment creaion abilities of tradable wages an a substatial scale, nor access to credit agricltureareaffectedbypre-exstngconditions facilities at rates of interest which make such and political economy considerations. Thus, a investment worthwhile. case can be made for policy modifiation or sup A key defining characteristic of smallplementation even though particular groups of holders is that they will be more disposed to smalholders are becoming better off under the supplement income thrugh selling labor, rather policy reform. Where land is already scarce and than through labor hre. What is relevant when the tendency towards concentration high, it may defining a smaluholder household, therefore, is be necessary to make a case for policy interven- the net quantity of labor hired, ie. the hours of dons designed directly to stimulate employment outside labor purchased by the household, minus creating opportunities within the mral informal the quantity of household labor sold off the farm sector. Food and input subsidies may have to be If this is expressed as a ratio, ie. as labor hired removed graduallyratherthanabruptly,butneed divided by labor sold, then an agricultural to be much more targeted towards nontradable household would have to register a value equal food producers. Thus, in general, some compro- to, or less than, one to be included in the mise needs to be struck between the optimum smallholdercategory.' policy and a reform that takes into account the 9.19 At the same time, because factor proporpolitical economic considerations and the anti- tions are generaly unchanging, and total labor poverty implications, use wfll largely determine the area of land holding, there will be a dose correlation between the Micro-meso analysis proportion of household to total labor and farm size, assuming a genera wfmity in houshoold hec aofsmitionoflnderhousdwols size and structure. Within the boundaries outlined above, therefore, the size of holding can 9.17 An Analysis Plan for smallholders s d provide a broad indicator of income. While some define smallholders at the outset and clarify how inforation is generally accessible, however, the theruapopulationcanbeusefulybrokendown delineation of smallholder groups by size of into wnstituent sub-groups for policy purposes. holding requires explit asmpto reguding This is the objective of the prest secdton. Te appropiate frm size ctegodes (irinormaton an sourcesof labor (familyorhired)and the applica- plot sizes per se is obtained in the SDA hite- 169

174 grated Survey). One metd for approading from fam and off-fani activities, as opposed to this issue Is to use average yield and price data to land size criteria, provides t first yadstick of esftimate hypothetical ionome levels per hectare asstion. based on 'typical cropping patterns. If a target 9.22 Thus, a usefd first disinction is betwee fam income is hypothesized, implying that all those households which achieve the bulk (mowe houldsewhchfailhtoreadizethsinemedt than 50%) of household income from off-faum concer in the form of policy assistance, then the activities (nrual laborers), and those whose priholting size necessary to produce this income mazy source of income is the fidly holding level can be calculated. The latter would provide (smallholders proper). 8hose faw in the latter the upward limit of the smauholder categomy as categoy can be divided further into a subsisdefined in ms of holding size. tenceoriented group (subsistence fames) and a 9.20 Thismethodofcategorizationisonlyvalld market-oriented group. The market-oriented In so far -is it is conducted for specific fam-ing groups grow tradable produce (see paragraphs systems (this may be approximated on the basis 9.26 to 9.34 below and Chapter Two on the defiof locational informadon available in the SQA rdtion of tradables and nontradables) for export surveys). Variations in the level of technology orimport-substitution (export crop producesand employed as well as In fte agro-dimatic condi- producersof tradable food). Thedefinitlon of the dions must also be taken into account if the area subsistence-oriented goup is again arbitrary (for criterion is to be usid effecfively. For example, exmple, no tradable commodities produced; or. smell-scale farmsoperatir.g witha high degree of less than 10% of total income derived from the capital Intensimty ay not always be usefully clas sale of tradable comnodities), but it B wonth sified as smallholders, 6 and the miin'mum area realling that all rural households are engaed in necessary provide to adequate" Income ay vary 6m fonn of exchange, including the sale of widely within countries due to different regional labor, and that there is no such phenomenon as a rain fall paterns and sol types. In addition, if fully self-suffcient household in autarky. there are regional variations in household size, 923 In addition to rural households involved holding size limits should be adjusted accord- principally (the peasants) or marglnally (the laingly. Thus, although land holding size provides borers) in agricultural production on the family a promising bases of clasfication, the Uimita- farm, there are those who dterive a lr proportions enunciated above make the importance of tion of their income from non-agricultural nonwage-labor a relatively better distinction criteria. wage laboractiviies. 9 The geat majority of the 921 Turning now to the classification of small- are likely to be invodved mainly In infonral scholders on the basis of the composition of their tor activities. Typically these activities involve Income, measures of the importance of wage la- artisanal or trading skills, and generate producbor relative to agricultural production activities tion of both consumption and simple Invesument and measures of the degree of market orientation goods from locally produced or acquired raw of households as producers of export crops and materials, to men locally expressed demands. food commodities are called for. ' An under- (Blacksmiths who produce and repair the modest standing of the extent of smallholders' depend- capital equipment essential to smallholder op. ence on the labor maket for a livelihood has emations are an example.) Growth within this been shown, above, to be important in our ana- sector (induced by agricultural production lyticalcontextbecauseofthepossibilityofashort- gowth) is of crucial importance if adjusent run real wage decine relative to agricultural measures are to succeed in stemming the tide of product prces as a result of structural adjust- rural-urban migration and, more broadly, if they ment. Indeed, market orientation will be mani- are to activate intersectoral growth linkages. festedbothvoluntarily(aswitnessedbythedeci Thus, as Indicted above, the following sion to grow particular crops) and involuntarily classification of the smallholder population can (as in the case of the household which is forced to be proposed (this is similar to proposas regardsell a high proportion of household labor to make ing the classification of the rural population made essential purchases). Furthermore, the basis of in Chapters Three and Ten in this Volume):" classificafon should be operationally simple and A. Export crop producers unambiguously undersadable and applcable B Producers of tradable food (import-substituuniformly to ad conditions. An arbitrary cut-off tion) point of the proporton of income originating C Rural laborers 170

175 D. Subsistece furiers in response to changes in world narket prices E. Rural non-faring non-wage informal sector (wbich are not under the control of national govworkem emmnts), and in response to a variety of measw 925 The aim of the above classification is to wres taen under adjustment prgms, pwartcgain an initial understading of the constraints ady those engendering movments in urrcy and opportluntit faing varioussegmnts of fte values or domestic price lev Given the volasmalholder population, later to be refined on the tlityof this type of dassification there are grounds basis of multivariate analysis using a household for suggesting that its usefulnss is limited. Inmodel capable of treating altmative production deed, when account is tken of the source of and consumption activities simultaneously. inputs used in the production of a particular Anlysts would start by setting up the above commnodity, ie. whether the inputs involved are categories of snialholders and studying between- tradables or nontradables, it becomes more senclass differences in incomes, expenditures and sible to speak in tem of the degree of tradabilstandard of livigindicators. Alotoflessonscan ity. be gained by using cross tabulations and fre 930 The degree of tradabiity is a hypothetical quency distnbutions as wel as means, standard concept reflecting trade patterns that would deviations and coeffidcents of variation of these energe in the absence of quanitative restricons "dependentr variables. in the product, factor and foreign exchange markets. In practice, however, the prices economic Cassij*aaion of products agents observe do not reflect domestic demand and supplycharacteristics or intenational prices RuralhouldsWparticularysmalholder Instead, prices are a result of govenment intrhouseholds,willoftenbesyinvolved ventions. Thus, the types and magnitudes of in the production of both tradables and non- tradable goods are likely to be different fiom the tradables. Neveritheess, a response eliciting a goods that are actualy traded. In sub-saharan shiftinemphasisbetweenthesecategoriesisiky Africa, it is common to observe parallul (bladck) to occur sooner or later in the wake of adjust- makets and smuggling, causg the edstce of men. Categ tion of agricultural commodi- more than one regionally differentiated price for ties into tradables and nontradables is therefore a the same product. The domestic price referred to *ecessaty first step in the analrysis. above sould thereore be re-defined to accom Theoretically a tradable commodity is one modate such country specific market intervenwhich wil be profitable to either import or ex- tion effects. The computation suggested above port given (a) the current level of world prices, needs to be done on the basis of he government (b) te exchange rate, (c) domestic price levels prke, the paralel market priceand someweighted and (d) tansport costs involved in intemal dis- avenge price (using quantities marketed through tribution. differnt channels as weights). As will be evi- 928 A formula can be constructed, based on dent from the discussion below, additional coitlese variables which wil demonstrate whether plications arise from differenc in tranwortathe commodity In question is an importable, ex- tion cost and the point of geographical refeence ponable or nontradable (see Chapter Two). Thus that further complicate the analysis. a commodity Is considered: 931 The tradable/nonntrdable dasfation of pmducts is usefud in tracing the impact of strucan importable if: P > P(l+t); tufral adjustment. A further distinction can also an exportable if: P e P*/(i.l; be made between protected and non-protected a nontradble If: P/(l0t) c P < l'fl.t), tradables, since the effects of liberalization, ie. the prices of Imports, and therefore of import Where P is the domesfic price, Pr is the wotld substitutes, canbe expected to fauilfurther followprice at the ofcal exchange rate and t is the rate ing adjustment If domesdc prices were previof bansport cost. ously Inflated by tariffs as well as overvalued exchange rates (Chapter Eight). At the same time 929 This classification is extemely sensitive to given the Impotiaine of iransport costs In these eac of the varables descibed. Apar from the calculations, thanalysisshould tkkeinto acount obvious influence of transport costs, commodi- the locational disparluet lit afstiiatlons which ties may cras from one dassfltioh to the other may mesult, and the onsequent ditferential Influ- In

176 ence which adjstanmt p sgran may have on houseolds ablity to respond to changes In the to product classification. general patternof incentives. Thus a comnodity Maret inbotion q*s can be an Importable in reo In the vicnity of a post but become a nontradable in an interior 935 Among the major economic phenomena town wher domestic transport coss make im- that dhracerize African economies is the pevaports non-viable. Those In remte area, therefore, will not have the same incentive to adapt sintervention of goveumnnts in product and factor markets. Diect and indirect intrventions therrcopplngoroutputpattern towardste types have been particularly wide-spead in agdculof commoditycruendy infavorelsewhem tural product markets, thus generating signifi- 932 These crcumstances can have consider- cant price dispaities across regions. Price disableimplicationsfortheoverall impactof adjust- parities are also due to a lack of infrastructur ment programs. In countries where commica- Policy reform and i ral deveopnt tins are very uneven, and where in consequence aim at unifying mrkets and reducing the disparlarge numbersof rural producersrenuin isolated ity in farm gate prices in different regions. To from international markets, a substantial propor- assess the progress made in this respect followtion of the population may not be able to take advantage of the potal benefits of adjustment. ing polcy refmn, it will be useful to collect and monitor changes in farm gate price. A f arm-ate This problem is compounded if, as a result of price series should be cntucted for different ependiture cuts and other measures, the pur- agro-1imatic reions wich can be compared with chas price of both inputs and necessary house- the average national producer price. Over-time hold puchase rises more than proportionately or regional divergences or nrwing of the gap In these remote areas. between these series can then serve as a measure 9.33 A tabuar preaton of the resdts which of the extent to which markets have been unified would illustrate the tradable/nontradable di- and the ectent to which macroeconoic polcies chotomy and the effect of transport costs on the have impacted regionally. The interest in this classifiction proess and the potential disadvan- exercise is in monitoring not only the absolute tes faced by producers in the more remote lo- levels, but also relative prices and the direction of cations, is useful. Based on this knowledge and price changes resulting from policy refm The ifrmation of local cropping patterns/output sures of these price data are published series pattems, a table can be constrced which illustrates the extent to which different housd and/or the SDA Community Survey. 936 Apart from giving an indication of the exgroups are involved in tradable or nontadable producton, and the extent to which this varies tent to which farm-gate prie are resonding to nmeconomicinitiatives,suchapriceseriescan by location. also be used to indicate regi diffenes in 934 It will be usefiu for poly purposes to the change of price signals. For example, any note the extent to which the different goups of regional difference in price changes could be athouseholdsareinvolved Inproductionof tradable tributed either to a concomitant change in the commodities. Althoughi, t is to be expected that, physical infrastructure and thus In communicain general, the proportion wfll be higher among tions and marketing costs, or to distrutional market oriented producers, the extent may vay argeents associated with institutional reacross regions as may the proportion in each fom. In general in Sub-Saharan Afica, costs categoryof housed The ifes tobedrawn associated with trnsportand distinbution account from this table relate to the ikelhood that the for a dispportionatepart of the final price, and greater the prportion of tradable commodities provide for very substatial locational dispwiin household producdon, the more potential ties. Any variations in market access as a result benefit households wil derive from the adjust- of govementexpentwtuepoacyorltitudtional nentprogram. Of coursethelnformationgained reform will therefore have a substt impact from a onepoint urvey wfll not take acaunt of on local prices. household responmesuch as changes in crop- 937 Thus price data can be used In two ways: ping pattem. To have an unds of the (a) to demonstrate the changes in local prices, dynmi effet IIIIIOf Suppresponse and therfore price incentves, which have ocis required. Before such an endeavor however curred folowing the intrducion of the adjusta_mft mnomiand real prie dffreil and ment program; and (b) to allow for a odmparbon levels of protction are a necessary supplement 172

177 of price Information across dfferent locationsand nreeare. The most commordy used are the thus to study the extent to which programs de- weighted price of purhaed inputs, to give the signed to improve the infrastructue, or con- input-output terms of trade; and the weighted verly cut-bacs In goven t spend have price of al purhased commodities to describe Influenced region price disparities, the ful net-barter terms of tade Both these 938 Real price series shod be calulated on measures are useful in the cunent catext, the the basis of the compfled nominal series. The fomertodescribehowpriceshiftshaveimpacted calculation of a time series of real output prices spe y on agricultual resource allocation, however requires an appropriate deflator de- and to higight the dfferences In this respect pending upon the question the analyst desires to between different groups of smaholders; the answer. Indices of consumer prices, prices of latter to describe the shifts in price patterns facnon-agricultural product, or prices of inputs or Ing ral households. intenational price can be used to reflect different 9.41 In calculaing the numerator, however, the aspects of the price study and to address differ- reseanchermustdecdewhethertheweightsused ent types of issues. It is unikely that locaized inaveragingprices(forestimationofapricefora price seriesareavallable for thelatterindices,but conposite agricultural produce) should refer to in as much as possible deflators reflecting rural quantities produced or to quantities marketed. conditions are the most appropriate. If these While it is true that It is the purchasing power of calculations are carried out for each comnodity, nmwketed output which will, in reality, deteran average for each household group can be esti- mine the volume of purhases, it is also likely mated, weighted by the sigfficance of each that for some groups of smafiholders, the quanticommodityintherespectiveproductionpatterns. ties marketed will be very smal and variable, The results will give some indication of changes being only that quantity "accidentally" surplus in local prices by household classification. to household requireents. It is pertinent, there An important real price series, which has fore, to attach a market value to auto-consumpremained as the focal point of policy coneern for tion and include it in the calculations. To be a longtme, is the terms of tradebetween agriul- considered, however, these commoditiesud tural and non-agricultural products. The agricul- be included in the composite index of pucased tural section of the SDA Integrated Survey col- goods appeaing in the denominator. lects data concer the prices of inputs and 9.42 The discussion with regard to nomnal outputs used in agricultural production and the pries, for differentiating the results by locafion prices of consumer goods purchased by the apply even more forcefully with regard to ters household. It thus provides the possibility of of trade data since prices of inputs and conompclcultingw terms of trade indices for each sub- ton goods are liy to vary as output prices. category of households. These data refer to one Terms of trade indices should thereore be calcupoint in time. Trends in these indices give an lated and presented for each household category indication of changes in the purchasing power of in each location. In making these calculations it households in the context of price sfts brought is useful to develop consumption profiles of the about by adustment proguras. Trends could be different households groups. This is especily established on the basis of repeated integated relevant with regard to monimtring changes in suveys. Repeat I's however, are not expected response to relative price movements. The conto be avaflable, except at considerable intervals. struction of a consumer price index by sub-cate- Butcommunitysurveyscouldberepeatedatlittle gory should also be undertake as illustrated in cost year-afer-year, as could the SDA Priority the Technical Appedix Survey (see Chapter Ten) The income and household tems of trade 9.40 There are a number of possible measumes are closely related. Coneptually, the two Indices of the tms of trade.' 1 Of inerest here are: (a) the aim to introduce the quantity changes that acnet barter, (b) the income and (c) the household company price changes. However, whie the interms of trade. The net barter terms of trade is a come terms of trade introduce the changes in the ratio of two welghted price series (P./P.) such as value of the agricultural produce, resulting from prices for agritural output and pries of non- both price and quantity changes, the household agicultural products (for Insc nc terms of trade extend the argument to include all goods). lhedenominatorcandescribeanumber income generati activities in the numerator, of price series depending on the objectives of the and the changes in the non-agritural products 173

178 in the denomator. By defiition, trefo, the dnge rates, the diere between the two Income terms of trade are a ratio of the index of providng an indication of leves of impidtt tathevalueofagriculturalsalestotheindexofnon- tion resulting from exchange rate distwtions. agriclural prices More formally, it is [(P.*Q/) Simlarly, the nominal rate of protection can be P.]. The household terms of trade, on the other compared with the official tariff rate, the differhand, describe the ratio of a weighted index of al encebeingatbuable to quanttatie estictions. prices associated with household income, fann The itepretation of the NPC is rather simple. In and wage income, and a weighted index of the general, NPC > 0 indicates that the product in prices of all expenditures. The household ter question receives protection and therefore Its of trade provides a complete index of the net domestic production is encouraged; NPC < 0 inimpact of price changes on household inome, dicates that the product is taxed and its domestic including imputed family labor income, production disouraged. The extent of resource 9.44 It is highly probable that regional wage pull(orpush)howeverdependsontheelasticties differntials will be altered by adjustment pro- nd the length of time policy reform has bee gamns, as areas most able to increase the supply alowed to impact on investment and resource of tradable commodities demonstrate an increase re-allocation. Issues in elas ities measurement in demand for labor. While a one-point survey and the dynamic effects of policy reform wil be will not necessanly capture this, it may be pos- discussed in the next section. sible to use other data sources to supplement the 9.47 In nodified form (Krueger, et al, 1988), survey information and thus gain Insight into the the NPC is a comprehensive concept for measurchanges taking place. In addition, the survey ing the impact of direct and indirect govement does present the opportunity to undertake analy- price interventions in the market of agricultural si of the iformal credit sector, since it includes products. I This measure takes into account the questions relating to the extent of los by source, effects of direct price interventions, quantitative and the rates of interest paid. restrictions, exohange rate distorns and terms 9X5 The analysis of noninal prices allows to of trade effects. The key elements of the measure _msure the extent to which regional price dis- are an estimate of the equilibriumexchange rate, parities have been alterd in response to adjust- and a disaggregation of the price of agricultural ment initiatives. But it does not ray much infor- and non-agricultural productsinto the traded and mation on the extent to which distortions, creat- non-traded components. The formulae using the inga wedge In domestic and inwnational pnces, new approach will be presented below but anahave been removed. The various terms of trade lysts are advised to refer to Knreger et al for the measures have also simlar weaknessesince they estimation of the equibrium exange rate. The are only comparisons of domestic prices. These, NPC can be computed as follows: again, do not inform us on the extent to which domestic price have been aligned with intern-a NRPT 1 (Pd/Pd)/(Pd*/Pm*)1-1 tional prices or the extent of adjustment required nrpd [(Pd- Pd/P.J/(Pd-/P1) in the future. The latter can be achieved through NRPI - [Pde/P,.)/(Pd /Pa )l -1 calculation of crop-specific nominal protection co.effidents (NO whih is basically the ratio of where: P&, (Pd-) is the international price at offidomesticpricestoborderpricesminusone. More cdal (equilibrium) exchange rate less tlansport formadly and in its simplest form NPC is com- and handling costs to farm site; P is the interputed as follows national price of nontraded good valued at equilibrium exchange rate; Pd, P. is as defined above; Pd NRPT is the total effect NPC; wrpd is the dit NFC= -- 1 effect NPC and NRPI is the indirect effect NPC Pe 9.48 The new measure breaks down the total NPC into a direct and an indirect effect resl_t where Pd is the actual donmstic ar-gate price from macroeconomic polcies. The comprehenand P 1 is the international price (in domestic cur- siveness of the modified measure is apparent in renc) at the farm-gate. several repects. For instance, the numerator In NRPIT is esseally the domestic barter ms of 9A6 NPC u requie separate calculations using trade discussed eadier. Simflarly, the sme equaboth official and shadow (or equilibrium) ex- tion reduces to the simple NPC described above 174

179 if the denominator in NRPT reduces to 1. The to go some way towards attributing causes to lihdiation of the NPC in its traditional or modi- effecl Given at the SDA surveys will provide fied form lies in the fact tat it does not acomunt cross-sectioal information and that some fam for ditrtiom arlsing from the price of imported management and me-se data are likly to be intemdiate Inputs. The latter effect can be cap- available frm other soures, more sophistcated tured using the concpt of the effective protec- forms of analysis are necessary if predkions are ton coeffcient (EPO. The advance to EPC is to be made concerning the futue impact of adbeyond the soope of this study, but it should be justsoent progrms on household incomnes and notd that due to the limited purhased input use productiotl in smalhholder agricultue in Africa, NPC and 9.53 Theanalylcalapproachmostappropniate EPC yield very close values. in these cirumtnces is to model household 9A9 The results of analyses on prines, terms of activity. Household models acodunt for the intrade, factor prices and NPCs can be readily pie- ter-eaiodshipsof production, consmtion and sented in tabular form A series of tables can be labor use. The use of agricultnual household constructed along the following ln of (a) out- models is wel documented' and there are a vaput price dhanges, (b) NPCs, (c) net barter terms riety of examples for reachers to draw upon. Of trade, (d) household tm of trade, (e) agricul. It is appropiate here to outline the main chactural wages and (f) wage tends for different smallholder categories. teristics of the model, discuss the validity of the underlying assumptions, and suggest the policy 9.50 A word of caution is appropriate regard- outcomes which it might be used to predicl A ing the use of price Indices. In many cirum- more technica discussion is presented in Box 9.1 stances the offical prices of consumer and in- (overleaf. vestmentgoods will have been largeyirrelevant 9.54 Household models combine many of the for many households, since supplies would not attributes of production functions, labor market have bee sfficent to meet demand at these equations and demand or expenditure functions. pries. Indeed, in many cases, local supplies may They are either recursive, in which case the dechave been entirely absent and offiial prices sions regarding production, labor supply and meaningles. The consequent ratining of goods conmption are made squentially, with the makets can distort analyses urness allowance is former influencing the latter, or they are nonmade for this phomenon. This caveat is par- separable, and each dedsion affects au the othtiuarly reevant with regad to expenditures on er Recursive models are most conmonly used, non-icturalproducts,agrculualhputsand partly because the assumption of sepability (ie. soial faciities such as health and education, that outcomes within different markets are not upplies of which have frequently disappeared interdependent) is reasonable under many from rural areas prior to adustment prorams custns, and partly because they are sim Pdreanalysisinitsvariousforms,whether pler and more straightforward to estmte. Conit nludes official as wel as parallel markets, is sequently, in many a recursive very much partal and static in approah. To be model is likely to produce usefud resuts. Alof value for prescriptive purposes, it should be supplemented by elasticity estmtes and sod though the model described in Box 9.1 is very simplified, it is easily adaptable to describe the proceed fromemni individual products to economic activities of a variety of non-farm understanding the effect of policy on the enter- households. Thus a representativehousehold for prise, that is, the smalmolder household. Thus, each smalwholder and infrmal sector group can the next section ains to incorporate the discus be modelled based on the mean observations in sion regrding prices into a household mdel eah categoy. If tim and data pemit, the exercise HousoldI model can be further disaregated by subdivi- sions based on location and smamolder classifications. 952 The methods of analysis and presenion 9.55 The household model, prsented in the described above are perfecdy adequate for ex Box, is based on the underlying premise of utility post descripion and explanation of a given situ- maximzation, where utility is derived from thre atio The objective of the Analysis Plans, how- goods or sets of goods; namely, retined houseever, is to make predictions concnng house hold production, purcsed cmmodities, and hold response to policy chan, and therefore leisur Utility Is amzed subject toa number 175

180 12 ~~~~~~~~~~~~~~~~~~~~~~7 Pill' d~i ihiihhii U~or

181 Therihhad side of ean (5) re s th m-pt tics of ll dte aglowu vabia Using the deumad for offul hwome( M. based an Oemakevahjeof household fam peduo (C.) fr htsao,r hwfcon can be fordmapedvtohdndeenterprinlidsrsfefna proftit may ddlnedas fiwa (amntngp maxblnb gfi f),lwe le0telft hand G. - p TD side represt the expwahu oan te three som of P,A.K.,T,DIW (1) houshidd utity. Given that the M lme egindon Mhe use of he model in etmatng th impact of polty Inhudes exprmion for te etme sourmas of lnoone, pwbt cm be de Ihe md by of fionownprodedklo,thesaleoflabwrandreltwa s,the kaowing example AASsmg ta the fm prodm s modcan beusedlonimtorhowhlbmfro ensouroe Ura4abIanenq depredation caueis Ipr ( to rimel liselytorespodtoag hdauinapoucyvaiabloet Tis wui have an Impact an both pmrdctn and owm pramer. umptlon.ftadioecohma eorywoulsggthat To fadtate esmation and policy lscum, the asove aut and marketd surplu wil bolb dae and comip. set of equatio have to be ewdtten in the reaed frn lion fal in rponse to the prie Iner Be h.ghe By sodoing the varbles that can be endgenoudy deter- outpt pre Impliessper prot, owver and weter mdned can be expressed as expwt hncdns of ao te pr t (ncme) Iy (for a normal good) inaeased on exogously given variabls Wt hand varlees) For farm consmpto hu, he impat on marketed surpls that purpos, fit substtute equato () nto equan (6) wil be ambipou& Given equation (5), the cosmpin and then equaion (6) Into equation ). 1U1 the ler- effed can be deduced frnt hand tem of theexpadedequation (5) to the ight nd sideandsettheequationequeltorarop /81nequationCI) i d -. + S. / M (BY/Z S) (9) ag follows where dg,/dp, 18 the agregate cnge of cmpoof L = U(G, Gd GG: D) *I)P.Q + (A, 1 K V: 8) -Gl the farmprodoe a whespe. toown prloee7 hetwo pat derivatives on the rgt-hand side resent the -P G..P 1 (T-L-GC,)-PPVE) (7) prte and proflit Gamm) exfc whuid ontlo ftto the total dhe In demand due to a dchage In own-prc. The _o vadables In equaton () are famlevnl Siar equatlons can be derived to predict a vaiety of retaned agmutupalproduc usedinhoushold aamp- diect and indred hages (aso the oresoaing des. tin (G), pu_cased commodides (G.). tiiom spent for lei- tdlies and c dro), andt alowing predle sure, demnd for labor (L) and demand for vria Inputs tim rerdng the Impact of any patcula policy sdk (V). At equrum, these vadls are em as a Iune ample, the u n for labor demand ard earlier, government restions In factor and therefore wages can be prediced through eud- product markeb affect the pries faced by eonmating labor deand and supply eastiles nondc decision-makers and condition private (3) The consequences for rural input and output production and consumpfion behavior. In the markets by estimating demand equatlons, eg. for definition of the mode although pries are aslocaly produced goods and services. sond to be determined exoeosy, allowance (4) The macro-economic impliations of polcy has not been made for the role of state maketing chang can be estimated through aggreption of agencies (SMA) and quiantitative restrictions. In household responses For examle, the conse- many countries, smauholder producers fie two quences for the foreign exchan position of an prices at the same ti. Pirst, SMA offer different Increase in tde martd surplus of an otewise prices fro those that would emep under marimported staple food. ketcdearvcondtions Furtmore, tse prices 9518 Current evidence wsugesthat In the on- are diffeetiated for consum and pdues text of b a Afica, it Is inaccurate to ac- and are concurntly suppl by requsicept some of the asmptioks made above re onrig from producs and/oratiordng to conardhg the operaton of the product and faclor su m This requies to modify the utility and markets Thus, the houseold model defined in production functions as wel as the time allocathe Box should be modied to acommodate the tion consrbaint country-specific circumstances under which 9.59 The utility function contains thuee sets of smalholder households operate. As dicussed consum goods and is defined on the assump- 177

182 tomthathouseholdscanpurchaseanyamountof of the labor market is called for in defnng te the set or subset of the products at the going me allocation consrant prie. In practice however it is fequently the 9.61 Inputusebysmallholderhou dsu other case that household choice is crumscibed by a than labor and land, is usually limitd. Very binding ration level. ResItitions have been scantznountsofferllizerandpestildesareused DmmmnnismanyAlsnlountieinrec ntyears by households, for example those who obtain paiulady for manufactured goods due to the these inputs as part of agricultural development foreign exchange shortages limiting the capacity prograns. The pries famners pay are often subto import munufachted consuner gdods and/or sidized and remowal of these subsdies may bean intermediate inputs and spare parts for domestic important component of policy reform Even if production. Thus, households have had no or famrs are willing to pay higher pries, howlimited choices other than amounts made avail- ever, the amount of inputs that they can use is able to them by state-run marketing or rral de- often determined ogenously, for istans, by velopment agencies. These are in some cases an extension agenl In such cases, an additional supplemented by purchases from parallel mar- equation or inequality should be included to rekets and smuggled products. But, most small- flect such constraints. This modification to the hc'-- households have not had such supple- simplified model in the Box 9.1 will make the meway sources. Therefore, in order to take ac- amount of fertilizer an additional right hand count of such binding restrictions, constraints variable in the reduced-form equation. In extreme need to be imposed on etotal amount of the set cases, fanrers operating outside of government or subset of products available for consumption. initiated programs do not use such Inputs at all. This fact limits the choices open to households. For those households who do not use any chemi- The utility function has to be vedefined accord- cal inputs, production in the short run is affected ingly. This issue is relevant in detemining the by the fixed land, capital and the variable factor right hand variables in the reduced-form equa- input of labor. Thus, specification of the productions of the household model. In cases wher tion function should reflect the drcumstances of binding quantitative rstrictions are important, different types of smallholder househods. the ration level enters on the right hand side 9.62 In cases where the functioning of markets together with the other exogenously determined for factors and products is restricted and prices at prices, and becomes an important policy instu- which resources are transacted are not market mrent. clearing prices, it is suggested to use vitual (or 9.60 Similarly, the assumption of perfect sub- shadow) prices From the foregoing discussion, stitution between family and hired labor is un- the need to estimate and use such shdow prioes tenable. Restrictions exist in the labor market is inescapable In the cases of the manufactured particularly In those countries that have at one goods, labor and variable Input markets. The time or another followed socalist paths of devel- exdstence of quantitative restrictions and other opment. Although structral adjustment will product and factor market imperfections can be have removed such restrictions and begun creat- incorporated in the stylized model in Box 9.1 In ing the environment for a free flow of resources either of the following two ways. First, in cases and labor in many cases, some vestiges of past where rationing is pervasive shadow prices that practices affecting the labor market ame likely to induc the same impact as te ration can be estiexist in some countrie. An assessment of the mated, and it will be appropriate to use those existence or absence of labor market constraints prices in the household model. In these cases, is necessamy even if restrictions have rendy been since prices are estimated endogenously, it is no completely removed. This h partcully useful longer possible to maintain that production dedif time series data are used for estimation. In sions are made independently of those concencountries with labor market restctions, house- ing leisure or consumption and the ecursive nahold labor use i8 limited to only own-productive ture of the model must be amended. An adaptaactivities with any shortwge rmaing unfulfilw tion of the household oudel to allow for simultaand surplus being only used to Increase leisure. neousdecision-makingtherforebecomesneces- This has an important bearing on the structure of sary. Although this situation can be handled the household model sin detemis whether using household modelling tnque te apthe model can be solved recursively. Thus, careful countrylevel assmen of the functioning propriate models are more omplex and the data rts more intensive." Second, in cases 178

183 where tt 1s difflcult to etmate such prices, both mwvement in the wage rate, which Is considered the prie at wich the product is ationed and the to be exogenous to the smakholer housebold. amount of raton enter in the reduced form equa- But, it will be more appropriate to solve the two dons These are important departures from the together partcularly as the analysis shifts fron standard household model described in the Box single houshd to the smallholder sectr as a and call for the modification of the standard whole and when labor maket conditions are exmowdel aned InmanyAfricacountries,thesmalholder 9.65 The simplified household model in Box sector operates in parallel with large-scalestate 9.1 together with the suggested modifications is agriculture, the latter using modn technobgy useful in Identifying the right hand variable, as and lrgely producing traded goods. The two shown in equation (8), that should appear in the sectors are either interdepen t and/or com- reduced form equatons. This approach is more petitive in factor and product markes Since comprehensive than ad hoc supply and factor policy measures directed towards the estate sec- demand specficatons. Nevertheless, to be usew tor affect the operations of the smadholder sec- ful for policy analysis and prerpon, the coeftor, it Is necessary to study the two sectors con- ficdents for each rhv need to be quantified. This cuetly. Consequently, identifying the type requires a correspondingly compreive data and taking their relatonship into account in the set covering the various attnbutes of the housedefinition of the smaliholder model, is indispen- hold. The Integrated Survey (IS) and the Comnsable In recent erience, the variable input munity Survey (CS), while not designed to fadili- (fertilizer, in particular) and the labor markets tate a disaggregated supply side analysis due to have demonstrated the existing intimute rela- the absence of questions aiming at relating inputs tlonshlp,thusstressingtheneedforajointanalysis to outputs, are adequate to estimate aggregate of the two production modes. Te relationship demand equations both for consumer goods and can be easily Illustrated by takdng ferlilizer as an intermediase inputs. example. Often smaliholder households receive 9.66 The data specifically related to smaluholder their fertilzer at subsidized pric while large- productive activities appear in the agriculture scale estates are subect to full-ost pricing. But, section of the SDA IS. The questionnaire will smallholder households may sell the fertilizer to collectdataonagriculturalassets(indludingland, the estates instead of using it on their own farms. lvestock and equipment), plot details, harvest A diversion of fertilizer, in full or in part, from and disposal of crops (including staples, field, smallholder households to the estates, creates an root and cash crops, and fruits), seasonauty of additionaldimension to the production and con- sales and purchases of selected crops, other sumption decisions the household makes. To the sources of agricultural revenue, agricltdural costs extent that fertilizer use on smaliholder farms is and expenses, processing of agricultural prodlower than it would be in the absence of these ucts and consumption of own produce. The data transactions, smallholder production is under- may be inappropriate for estimating annual agmined. But, the loss in production is compen- gregpte crop prduction, yield levels, input-outsated by an income bansfer generated in retum. put ratios, crop-specific production and supply Thus, the analysis of policy refonn on smalholder functions. It is however useful to estimate aggrehousehold's production and consumption deci- gate farmincome fromdifferentcrops, aggregate sions should take into account not only the ef- factor and input demand, volume and value of fects of policies direltly addressed to the small- sales and purchases from different channels. holder sector but also Oteir effect on estate fertil The lack of dsrtion and the failure izer deivery. to capure oss-price elasticities howevisikely 9.64 Similar inteions exist in the labor mar- to pose difficulties in attempts to understand the ket. For ance, an increase in the eative price effects of crop spedfic policy reforms. To introbetween the estate and smallholder produce, in- duce these latter effects adequately, the SDA data duces an upward pressure on labor demand and nstuuments will have to be supplemented by the rural wage rate. his is Ukely to create a flow farm management data or analysis based on time of labor from the smaliholder to the estate sector series data if the latter already exist in a fairly which may ultimately result in labor shortages in disaggregated form. Caution should however be the smaliholder sector and affect its production exercised in using time series data since the variadversely. This effect is indeed captured by the ables that are assumed to be fixed in the short- 179

184 run In the production function ate no more so in sea t keep in mind th bmad context in fte long run. Similuay, as the data are generated which te respons to polic changes am tking from economic conditions at particular times in ple. Gi the often raial natwe of the prothe past, any change In economdc policy regme gnis,andthelklhodisthatteycaurypottal over the period will affect the data and require a implcatons for almost all aspects of the socal corres ng dificat In the specification and econodc lfe of many rur houselds t Is and estimation of the equatiors This, therefore imta that aaen is not wuduly focussed requires a clear u t of the economic on a minority of variables, such that only a parregime(s) for the period to which the data belong. tial view Is obtained The need to re-si the model to reflect 9.70 The first set of policy implfcati to be realities of specific couries and t compromse drawn stems from the ramifications of adpustresuiting from data limdtations is compounded ment progams for householdss capay to gtby technical difficulties in aggregation and pa- erate inome. The extent to which access to asrameter estimation. Firsy, modeling based on a sets, inputs and employment opportuities is representativehousehold risky due to the fail- beiginfluenedbypolicydcangeswillultimately ure of such a unitay classfication to caphue the be a key determinant of succs or failure, and effectof incometransferacrosshouseholds. Such thendoreprovidegudancetofutuepolicy. Thus, shortomings are also likely to or at the aggre- changes in access to credit facilities, the availabilgate smalholder level, and in using time-series ity and access td land and devdopments rearddata,sinceforagivenlevelofincomeitsdistribu- ing rural employment opporhnities need to be don to one or the other segment of the small- monitored and appropriate polies initiated. holdersectorproducessignlficar 4 departurethan 971 The second focus of attention has to adwould be the case based on a representative dress the implcations of policy changes for fam household. Secondly, given the crossection profit. The outcoms in this respect may well nature of the data, It coud be difficult to capture vary considerably among the different smallthe effect of price changes on the endogenous holder categories, depending on the extent to variables of policy concern if diffeence in price which (a) the household is engaged in the proacross regions or seasons is Insignificant. This is duction of tradable and nontradable products, true in the case of pan-territorial pricing. In those (b) the existnce or absence of competition in countries where pan-territorial pricing has been export and domesic markets (the latter applying the rle and fully enforced, the absence of vari- to inport substituting products), (c) output price adon in prie across households, regions or sea- adjustnt and the ability to switch to more sons will make it almost impossle to estimate remunerative products, (d) use of non-labor purprice-related elasticities. But, the faiure of en- chased inputs and (e) the extent to which the forcement and, in parallel, the segmentation of price of purchased inputs has been affected by product and factor mawkets has in many coun- subsidy removal and other policy refom. It is tries produced highly differentiated regional possible that a locational dichotomy emerges, prices. This latter development will perhaps ac- between reglons where tradable producdon, parcord a proming opporhunity to adequately cap- ticuhrly of exports, is feasible, and those regions ture price effects in a cos-section data. The sitable only for production of nontradables or third issue reiates to prices. In the stylized small- non-procted Inports. It is equally possible that holder household model, the household is as- a proportion of households may be bypassed by sumed to be a price-taker, as mentioned earlier. policy refrm In either ca reseanhers are en- If households are aggreated regionally or sec- couraged to suggest policy nodifkations which torally, price cannot anymore be considered ex- wouldenhranefarmprofitanongdisadvantaged ogenous. Istead, it becomm endogenous to the groups. sector. Thus, country-level analysts need to be 9.72 Furtmor, a more complete indication aware of theimplicationsof the varius assump- of the impact of adustment pgrams is protions they make for parmeter estimation. vided by t of modifications in household expenditure behavior. To this effect, indica- Policyperpectiv tors combining information on inome generation with data on consumer prices and expendi In review the policy implications of tue profiles, provide a cmprehensive mas of adjustment progrms, it is important for re- evaluating the Inpact of the programs on the ISO

185 various household cltegdries. An essential part dchsed Inputs are avaiable from thesda IntgatedSuvey.) of nyevaluaton must also Inude an attept to 7. As i evident fom the foregtng and subsequet sacestmate the I iatlmons of the program for ac- dos the c malihds at n rather lowdy ap. cess to economdc and sodal infrastructure. Easy plied. In some section the dscussi ompn s e runa and affordable accs to makets and sodal serv- at lare induding farm labors livestoci rser ices such as education and health faciities con*i- and non-farming non-wage rural inml labor under the tute an essential component of household incme, geal rubric of smaho. Thus, the sm a discus- Whether such access is Impving or deteriorat- sin whould be put In a much boader cntex as and whe ing, researchers are encouraged to inepoate neceary. such normation in iheir ovel appaisal of 8. Uvestock plays an Integal part of the smaholder prgams implications. eonomy and is, in e areas, the prindpi focis of pro When interpretg the results of the SDA dudian Whep b caobedirdasholdas surveys, therore, researces are encouraged to in the conventoala sense, they are frequently among the take acount of the social and econoidc context pooret of ral populatis and possm many of the me which shapes household welfare. Concentration broad dhartic with respecto the exdguous and fra on the more easly quantifiable indicators to the ile resour b and low leves of _amamplion. hus, load exclusion of others may result in nmsleading con- researchers must answer wther pastoralistshould be clusions. The noton of household income must eatd as a separate group, or be subsumd under one or be intpreted in its broadest sense if the ranifi- more of the ot categries cations of adjustment programs are to be fully 9. Itisiely that the majority of url hoeholds would understood. meet the income criterion for incluston among therurl poor. Nevertheless, there wi bea ndnority, such as professional, Notes govrm t employe and those invdved In large disbu tion and maketing activities, with which the SDA Prom 1. Thls sectin Is based on a _mpanion volume, World wil not be concerned. Bank 1990: ). 1. Thellstlstentatlve,andeard smaywishtoextnd 2. Land eorm have not In geneal been assodated with arreducethenunerofgroupslnvolved. Itmaybepetinent adjustmentlending beouse fidth meframehas uuybeen In some contr to caeate a sub-ctegory of smalholders la tan the adjuament program An exception Is their acoding to e ic orgn espeiy When pa lar ethic Inorpoain In SAL a g with _ Kenya, with the Inten- groups have restricted acm to land. tin of gvng land ighb tosquttrs. I1. Fwrausdfuemplrlareviewofapproprate tchnques 3 Altougb there are pocket of land scardty In Sub- for calulating terms of trade, reseaes Saharan Afic the major mnsbtrt to production (as statd FAO, 1966 to are referred in paagrphs 9.01 to 9.01) is not acc to land, but the 12. Two arlldes which lhutrate well both the usefulness availbityof labor. Anyuseldeition of thesmalholder and poten pitalls of NPCs are M.J. Westake, 1987; and populat mst n tharore take into aomt this overding W. Jaeger and C H 9umphrey contai to household Income 3 The reoent development in the mncept and its empiri- 4. There is a danger In this definiio however, that the calappeliaosinmanydevdopingcountiesrlaudingchte labor nees of sme grou8 of very poor households are divoire, Chana and Zambia in Sub-Sharan Africa, can be Ignored. Frezape,IfthehouhAtdhead isa woman and found inanneo. KIruegerMauriceScifffandAlbertoValdes, land reparation invoivesasubtanil amount of heavy wok which she is unabe to undertake. Analyst should be cognl- 14. Fora detailed outline of a basic agricdthal household znt of this pity. model, and examples of the use of these models in practce, 5. Atwgetincmewouldbebasedon teestimated value the reser is erredto LSngh, L Squle,J1-Stauss (eda), of an eential household consum profile Analyss 1!11K should use their knowledge of loal cdons and assump- I1. It is possible to attibute a value to all avauable time, tim regading per capita calorie and nutient requiemnt sic time spent In Idsmeor production of househod goods (coupled with average family sie data obseved in the SDA such as home-maing and child nurturing, has an opportusuvey In derivingsudi an estiate. nity cost in terms of the income foregone through not ubliz- 6. Thus,a fa(mer operatig a smaarea and employing ing that hour in farm (or off-arm enteprise) production or little non-household labr, but at he same time using mthi- the sale of wage labor. ods Invlving a high degree of _edwalzatlon and/or the 16 It is noteworthy that weather variability and its preextnsve use of purhased dml Inputs, is unliky to be deability are Inportant consderao In famersm chice of amongt the nrral poor. (Data on fam equipment and pur- cropping patt In as mch as price a infrastructural 181

186 -: li 1: -i1. jii' 1.X, *.0 7~~iI

187 .1'~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I I- An analysis plan for the SDA pri'ori'ty survey Intrdction over time is different from analyzng cross-seton variations The fist useof the PS in a adun The SDA Priority Survey is part of the tiy wlllnevitablybeforthelatterpupose,hence hiformation system proposed under the SDA thisreevestebulkof ouratlentmonthischap- Prgam and is aimed atproviding, inparticipat- ter. ing countries, the empiical basis to assess the This Plan is differet from the others in social Impact of structural adjustment prograns. tis Volume in three respects. The SDA information system is described in () It covers all sectors.and taget groips, in con- Chapter One of this Volume and, in greater de- trast to, for eample, the plans on health and til,in Wotid Bank (1990). A basic pmise is that education. This stemsfrom the limited data conany single survey woud not be able to meet all teit of the PS. The survq, however, should not data requirements, but that instead a hdiarchical be egarded as a "mini version of the IS. Ixdeed, and integrated data system is needed. The sys- the PS offers its own analytic possblities, e.g., tem would need to intlude surveys capable of those based on the inlusion of direct retospecproviding the extensive data required for inepth tive question on change. Hence the analysis policy analysis, as well as the more urgent data preseted here is not simply a sub-set of the IS needs for short-term policy decisions, partlcu- analyses - which is why there is a separate PS laity on trends in "dependent variables As Analyss Pla stated in Chapter One, the Priority Survey (PS) (i) The analysis is applied only a; the microaims to identify target groups for social aspects level the orientation of the PS is towards providof adunt programs and to establish for such ing a set of indicators of how strucbtl adjustgroups key indicators on welfe and basic needs ment may affect the seected taget grups, and fulfmenl these are at the micro4level When combined This Analysis Plan focuses on the PS in its with a CS, the PS can of course be used to study primary role of idetfyin Wget gtoups and meso-wmicro links, in which case the nite of establshing for them indicators through which these links is much the same as thatdescribed in theimpactofadjustmentcanberecorded. Wbhle the IS analysis plans (although the amount of we dsall occasionaly refer to the importance of feasible analysis wm be less than in the case of specific tables for monitoring purposes, in gen- the ICS combination). eral we shall not discuss in this document the (Oii) Te analysis is nsly confined to tabulaanalysis of PS data for monitorng. This is be- tins. This also stems from the restricted data cause the PS questionnaire may be somewhat ontent of theps, which pemits thecnistction different in its mon$toring version and because of a set of useful indicators and their cross-tabuthe statistical analysis of changes in indicators lation with relevant target group identifiers. t i I - 1 ~~~~~~~~~~ 83

188 However, by and large, the varables needed to put more emphasis on local polcy prionties. explin these indicators, in a modelling sense, ae The in-depth analysis made possible by is*ing in the PS. They are presnt in the IS as the IS will, quite reasonably, take a significant explanatory variables to model household be. amount of me, which is acceptable given its havior. focus on medium- and long-term planning and 1004 PS analysis is particularly questonnaire- policy design. In contrast, the PS information is dependent - more so ftan IS analysis - in that aimed at policy interventions which are needed even mnr modifications will alter the analytic in the near term, and the results must therefore potential of the instrument. This Analysis Plan beavallablewitlunweeksof data collection. This must therefore be read in dose conjunction with is of course one of the sever aspects of the the PS annotated questionnaire (World Bank, complemenary roles played by IS and PS in the forthcoming 1990a). The latter discusses a num- SDA infornation syste5m ber of choices and trade-offs with respect to ques This near-term focus of the PS immeditionnaire content and field application and incor- ately eliminates several items from the data conporates the experience of a PS field test conducted tent, such as the caphtring of seasonality, full in Ghana in early Since the questionnaire income and expenditure accounting and interon which this Analysis Plan is based is a proto- views with all adult household members (all of type version, the analysis proposed here is also these features figure of course in the IS). In addiprototypical, and both will be adapted and tion, for the PS to fulf its role properly, the amended to take into consideration country situ- information collected must be easily measurable ations and priorities and obtainable from one, or at most two, respon Thetextbelowwillnotdiscussall possible dents per household. This is necessary to keep tabulations, but focus on a few key tables on each the total Interview time per household to within topic by way of examples of how to construct and one hour, by avoiding time-consuming re-visits Interpret tables. In the next section the data con- to contact different menm An inunediate tent of the PS is discussed in more detail. The implication is that the focus of theps wiu have to issue of socio-economic classification and its rela- be on household-level variables. Information at tion to the target group concept is explored in theindividual level generallyrequiresinterviews pargaphs through Paragraphs with all the individuals concemed (as is the practhrough 1092 constitute the Analysis Plan proper, tice in the IS), since proxy responses are usually covering basic needs, employment, and income not reliable. Ideally, the PS questions are adand assets. A concluding section rounds off the dressed to the head of household only - but on text occasion the spouse wil serve as back-up or supplneentary respondent. Data content From an analytic point of view, the variables in the PS must be proxies for direct welfare The overriding characteristic of the PS in msurement and be sensitive to changes in the the SDA information system is its ability to pro- economy. The need for proxies stems from the vide rapid information on key variables to policy fact that the complete recording of income and makers. This determines the data content and expenditures cannot be done in the time span of field organization of the survey. In contrast to one hour (and would require interviews with the LS, in which data collection is spread over one many household members at any rate). The seyear, the PS is to be conducted over a 2-3 month lection of such proxy variables isin fact the single period at maximum. In order to pemit sufficient most importnt (and difficult) task in the design sample sizes within such a short period (and of a PS. Moreover, the proxies must be chosen so keep the number of enumeration teams reason- that they are able to reflect the impact of adjustable), interview time must be kept short - the ment or other macro-policies. For example, one goal is a maximum of one hour per household. might argue that the literacy rate in a country is The fildd test of the PS in Ghana has indicated one of the nwst Important indicators of advances that a country adaptation of the prototype ques- in human capital accumulation, and that literacy tionnale on which this Analysis Plan is based is an important ingredient of personal welfare. can be admnistered in about 30 minutes per However, the literacy rate has a significant mohousehold. There is thus scope for expanding mentum component, and will not change much the content of the PS if countries wish to do so to in the short run, say, as a result of a drop in 184

189 educational expenditures. On the other hand, annotation to the questionnaire (World Bank, the student drop-out rate will quiddy respond to forhfing, 1990a). he explanation of the use changes in the monetaryand/oropportunitycost of these data fonn the task of this documnt. In of scholing that can occur under adjustment. the meantime, Box 10.1 sunmarizes the data The PS should not therefore, in this example, content of the PS. measure the literacy rate, but focus on such variables as primary school enrolment and recent Target groups and socdoeconomic groups changes in it Relevant indicators can then be constructed, such as the drop-out rate among The primary objective of the PS is to idenprimary sdhool children and the (mis)match be- tify target groups for social policy interventions tween age and grade attended. Such indicators will readily reflect undesirable effects from adjustment in the area of education. Similarly, as far as farm enterprises are concerned, the PS can- Box 101: PS data content not attempt to construct the income from such I.Householdroster(lstof allde fthouseholdmemactivities (this is an important role of the IS) but bers)- must focus on those aspects of operating a fann * deomy-phy. reatn to head of Jhold, ag, which can be influenced by adjustent, e.g., price a" changes of crops, price increases and/or changes * educitoarmunvtahttdanceducaloa1addevein availability of imported fertilizer and the like. ment A key functionof the PS is to record unde- * healt h:eet hlth o lton and reldated x- sirable changes in key indicators. In its monitor- pe mj ing role, this can be done by comparing results 2Houag and amnttes: from successive surveys. However, even when * oapanq st*u and dange undertaken for a first time, the PS has the poten- * awesstowaterndfuelandchangeinacoes tial to record change directly. This is achieved by &Empyformheadofseholdanddspoweny) direct retrospective quesdoning on topics for * prsfce in houshd which the respondent can reasonably be expected n oan wok sas Industry, wage to recall the situation of, say, a year ago. This second )Con *previous occupation function is in fact more imponrtant in the PS than 4.Miaffm in the IS, and is one of the reasons why the PS is * recet migaodn of howehod nota "mini-is and permits analysisbeyond what * work-reated miaton of hoehold membes is possible with the IS. Such etmospective ques- S. AptaslentepseOndudnglivestdandfishtions are less useful in the IS because its inter- oes): views are spread over 12 monfts, which makes * ownwdpof land the comparison of two points in time more diffl- * fsmahipsprosaesaetle g cult. In the PS, direct questions on change are * e of fet co, tensio servis incorporated, regarding occupation, agicultural. jingof wdwscandcanges in hinws production, crop area, hiring of workers in house- * maiivestoc an dhange in stoc hold enterprises, access to housing amenities, 6. Non-farm entepises (up to 3per househodk) migration, expendires, income and assets. The * grm nosme latter deserve special attention. While the amount * type, duratin of operadi and value of household assets are important wel- * persen responsible fare indicators (and are collected in the IS), 7. Hofsehold Wlod:tana changes In assets play a special role in evaluating * h and drectin of chnge for S key tnds in household welfare. Decreases in par- foods and mao rtegoissof non-food items ticular assets often reflect distress sales and may & hncome bysouron reflectincreasingdifficultiesingeneratinghouse- * Income and drectin of hange of major souroes hold income- The PS therefore has a set of ques- of lfof tions on changes in the past year in the 9. Asss: * owmnhlpand dca in owneip of majoprohousehld's holding of livestock, housing, pro- dueive as d adumdm ducuve assets and household durables. 1ao Mn q A detailed discusion and justifcation of. weitandheiofchldreno nthsof age what is inuded in the PS can be found in the I 185

190 in the contedt of adjusuent. This means that the is offered; PS nmst contain the necesay identifiers to define such groups from a policy-relevant perspec- 1. rural - export-oriented farmers (medium and tive. The first task of PS analysis is hence to set large holdings) up a classification of socio-economic groups 2. rural - export-oriented smaliholders (SEGs). These groups are meant to be as ho- 3. rural - food/subsistence farmws mogenous as possible from the point of view of 4. rural - pastoralists the impact ofadjusent on their welare through 5. rural - landless agricultural workers the production and income-generation mocha- 6. rural - non-agricultural workers nisms of the economy. Moreover, such groups 7. urban - skilled public sector employees must be defined along dimensions which can 8. urban - unskilled public sector employees become criteria for the implementation and tar- 9. urban - formal private sector workers geting of government policy. 10. urban - informal private sector workers A priori, two major criteria can be put 11. urban - inactive or unemployed. forward to define SEGs in that perspective: (i) Location: this refers in first instance to the ur An important variant of this classification ban/nual distinction. Where possible each can would be to dstnguish between the capital city be broken down further, e.g., rural areas can be and other urban areas in the country. This is splitbyagronomicorecologicalzones,andwithin warranted if, as is the case in many African counurban areas the capital city can be distinguishwd. tries, the population of the capital city is a mul- It is generaly not desirable to present results ac- tiple of that of the next lagest city. The capital cordingtoadministrativeentities(provinces,dis- city then often displays many economic charactricts) because these rarely represent economic teristics as well as problems which are unique to entities. Also, it is inadvisable to use "survey it. For example, the concentration of economic regions," i.e., the territories covered by the sur- activities (government, export businesses) may vey teams, becuse this holds the danger that nman that the capital's economy will respond enumerator effects unduly affect comparisons very differently to adjustment than other urban across regions. areas. However, splitting the urban SEGs in this 0ii) Assets/sources of income: this is clearly the fashion may lead to a smal number of sample prime classfying variable if economic homoge- observations per SEG, unless the sample size of neity is strived for, since this is the main mecha- the PS is very large. In general, a trade-off willbe nism through which adjustment affects house- necesary by aggregating along another dimenhold welfare. In rural areas, the classification can sion. For example, instead of simply doubling be made according to whether the household the number of urban SEGs in the above example owns a farm (and the size of it and its major crop classification from 5 to 10, one can propose an or type of crops (export vers subsistence crops). increase to only 6 groups by using the following In some cases the institutional arrangement un- scheme: der which the farm is operated can be part of the classification (e.g., share-crpping). In urban ar- 7. urban - capital city - public sector employees eas, the work status of the household head and/ 8. urban - capital city - formal private sector or the sector providing the household's muin worers source of income can be used (e.g., government, 9. urban - capital city - informal private sector parastatal enterprises, formal private sector, in- workers formal private sector). 10. urban - other cities - formal sector workers 1014 In some cases the classfication can further (public and private) be refined by certain sociological criteria such as 11. urban - other cities - informal sector workers ethnicity, religion and language, in countries 12. urban - inactive or unemployed where this is deemed appropriate and policy relevant. As a general point it needs to be empha It is recommended that the SEG flassificasized that the SEG classification will be country- tion be detemined a priori, at the stage of survey specific and needs to be built so that it can reflect design, for the following two reasons: (i) the PS local conditions and policy issues. Tlerefore, as sample can be made more efficient if it is stratian illustration-awltugh perhaps one with fairly fied along some of the retained SEG criteria; and wide applicability - the following classification (ii) a clear Idea on the SEG classification is neces- 186

191 sary to determine tihe predse data content of the nomic group(s) to which th belong. PS Tlhe are thee ypes of relation possible It is of course possible, and in most cases between target groups and soco-economic likely, that as a result of actual analysis, the clas- groups: sification will be amended because some groups (i) a taret group equals a SEC (eg., export-oriare found to be very simflar to one another, ard eted smalholders);(ii a target group isa part of can thus be aggregated, and/or others are found a SEG (e.g., laid-off, unskilled public sector to be too herognous and need to be split. employees); and (iii) a target group overlaps dif In the PS data base, several sections need ferent SEGs (e.g., fenale-headed households). to be relied upon to construct the SEG classifica- An important task of PS (as well as IS) analysis, tion: therefore, is to show the inddence of target groups * the urban/rural infonnation, as well as the within the different SEGs. This is essential to possible distnction between the capital city and assess the efficiency and leakages of prwams offer urban areas, is derived from the question- which are btgeted according to SEG criteria. This nare cover sheet which identifies the cluster; task (in addition to the display of the overall * the classification of different types of famers population distribution across SEGs) is the first usesthe crop and ownership of land information step in PS analysis, to which we now turn. The in the PS; main information source is the household roster. * thediitincton between agricultural and non Frame 10.1 shows the distribution of indiagricultural workers In rural areas, as well as the viduals by age and sex across SEGs, as well as the breakdown of utban workers, is derived from the total population distribution. This is a basic occupation information in the PS. demographic table which can be chedced with Once the basic SEG classification has been other sources of infonmation if they are available established, It becomes a mahor dimcnsion of PS (such as a recent census) to assess survey reliabilanalysis. The majority of tables will have SEG as ity. The details of the age/sex distribution can a dimenson and the basic description of the reveal important differences in demographic sample population can usefully be done by SEC paltterns for example, some SEGs may have less (see below). than the average fihare of adult males, perhaps as 1020 Prior to starting tabulations, however, it is a result of migralliin. important to clarify the relationship between tar Frame 10.2 (overleaf) labulates, at the get groups and SEGs. The way they are defined, household level, the relative importance of the SEGs form a mutually exclusive and exhaustive SEGs In the overall population. Thi3 is the key clasfication of households in society. They are table to show the correspondence between tarm fairly homogenous intemally with respect to groupsandsegs. Forexample,iffemale-headed major econoc parm rs. Target groups, on households are a target group, their distribution the other hand, are selected sub-groups of the is shown in line 2 (which sums horizontally to population of concern to policy makers, and are 100%). Une 3 then shows the percentage of fedefined along various economic and non-ec- male-headed households within each SEG, ie., nomic criteria. Tiuget groups can overlap with the incidence of the target group within the SEGs one another so that one individual or household can belong to several target groups (or, of course, to none at all). For example, the poor, female- Frame 1O.1 The distribution of the populaheaded households, malnourished children, tion by age, sex and SEG smallholders, coffee farmers, etc., can be target (p,,t groups. Tart groups need not necessary be defined at the household level (eg., malnour- Age See SECG 1...SEG11 Totl ished children) and can be quite heterogenous M (e.g., the poor). In many instances the ageting p of polices needs to rely on sodo-economic crite M rla which make ft possible to identi in practice F the target group. To identify the poor, for ex- * ample, one might examine locaftion, lack of as- 65+ M sets, ertai activities, etc. In other words, the Total P 10%_.1 o% poor are identified in relation to the socio-eco - 1OD OD_ 1 187

192 Psae 10.2: Distribution of households armss of halth and education activities, and threby to SEGs affect the demand for these services. As a moti- (puswis) etoring instrument, the priority survey will provide policy makers with the information to judge SEG t_...seg I Tthe extent to which the polky reforms that are AUl huwbehds 100% inplemented are having real effects on the basic Femak4msedhownhokb 1aOs needs status of the various regions and socio- Pert o fema1eaeded economic groups. The tabulations presented here h_dsehol are disaggregated bysocdo-econonic group. The socio-economic disaggregation is the main feature of the PS data analysis. These tabulations (togeer with male-headed households, this suns may also be disaggregated by region should polto 10% vertically). The table thus indicates the icy priorities suggest this would be useful. extent to which the target group is concentrated The PS focuses on four main areas of basic in a few SEGs and to which SEGs any interven- needs: education, health, nutrition and housing. dions need to be targeted in order to minimize Each of these is considered in turn. leakage. Education Basic needs 1026 One of the problems with education indi As we have already indicated, the SDA cators is that many of them have a momentum Priority Survey places some emphasis on basic which changes ody gaduafly. The literacy and reeds indicators and obtains information on the nunwacy rates are examples of such variables. education and health status of all members of the These are generally 'outcome" variables - the household. Indicators of nutrition and housing outputs of the education stem. The main conare also obtained in the PS. The reasons for ths cern of the survey is with the inputs of the educaare evident. First, these variables are important tion syste, tracing the numbers of the schoolndxcatorsofwelfare,traingbothshort-andlong- age population cutly receiving educatin. The run effects. Indications of morbidity and malnu- basic indicator which underpins most of the edutrition, for example, are evidence of immediate cation analysis of the PS is the enrolment rate. stess among the populationl But basic needs This is defined as the number of hldren attendindicators also suggest long-run implications, ing school (in the PS the definition is "regularly simply because they are important inputs into attendingschoolduringthecurrentschoolyeart) human capital, and thus influence incomear divided by the total nunber of dhildren of sdhool ing capacity (and welfare) in the future. SeC age. Enrolment rates should be calculated for ondly, they are relatively easy to measure (at various levelsofschooling,highlightingprimany least as compared with income- and expendi- (aged 5-11) and secondary (12-15) education. ture-based indicators). We can be reasonably Undess there are specil policy considerations confldent that the household head (or respon- which suggest otherwise, the analysis of the PS sible respondent) will be aware of the education should not deal with terftary levels. Only a small status of each household member and the health proportion of the population wil be attending at conltacs each has mode. Thirdly, structural ad- this level, and the majority of these will be from jusemenl typically Involves fiscal policy reforms, better-off households. Moreover, it is likely that which week to reduce the fiscal deficit through the number of observations in some of the cells cuts in governmn expenditure. In many coun- will be too low to draw meaningul Winfrec tries, expenditures in the heallh, education and These rates should be calculated for each sodohousing sectors are cut, whilet at the, sames time, economic group (and for each region if this is subidies of important food hems are rodume (or considered important for policy purposes). These removed entirely). Theseausterity meaunm,aea calculations would yield tables following the seriously impinge on the acmess of households lto etiture of prame basic needs services, and thereby adversely affim 11).27 Compaulsons between the socio-economic household members. At the same time, chanes gmups will not only be interesting in and of in the income-earning opportunities of house- themselvet flkv will also provide some initial holds are likely t change the opportnmi a clues about the underlying fatrs that might be 188

193 Frame 10.3: School enrolment rates by level of schoolin& sex and SEG Frame 1Q4 Educational drop-out rates by soclo-economicgroup SE. G Tl SEG... SEGO1_._W..._5EC1Tot 11 Priumry enrlmet rates Pmary-level M dropout rates: F M T F Secondary ewolment rate T M Secondary-level F drput rates; T M FT responsible for the variations in ermllent. For example, differences between rural- and urbanbased hoseholds mught suggst variations in the availability of schooling. On the ot hand, Frame 105: Percentage of cide not cu low levels of enrolment among farming house- rlgtic astn school forfinani and holds may well indicate high returns to child listicreasons, byleve of schoingand labor in the agricultural enterprise. The differ- socdo-economic goup ences in household income between SEGs will. also explain some of the variation in enrtolment. SEG 1-...SEG 11 Tot This is not to suggest that such differences wil in any way prove such causal links, but they may Puwy educti provide some intial guidance to policy makers. udalneswaints Distance from school Some preliminary multivariate analysis may be swonda education feasible with the PS data, and this is discussed Finamialcostaints brefly below. Notice that such tabulationshould Distance from school distinguish the sex of the children. In this way, the analyst ca gain some understanding of any gender bias in swhool enrolment among the sodo-economic groups. attendance, the reasons for non-attendance are One of the problems with the enrolment obtained. Of particular interest to the sodal dirate is that it may not be a sufficiently sensitive mensions of structural adjustment ae the rendicator to capture the effects of recent policy sponses which report 'finance' and 'distance to changes. It should therefore be supplement the schoor as reasons for non-attedance. Both with some analysis of recent drop-out rates from are fctors which are affected by govement the education system. The PS obtains informa- policy, and which can potentially be corected (at tion on al individuals in the household who have least given time). Thus Fram 103 and IOA left school during the current year, and the rea- should be supplemented with tabulations indisons for so doing. This is certain to be a more cating the main reasons for non-attendance -as sensitive indicator than the enrolment rate. Thus, for example in Frame This table can be a tabulation along the Lines of Frame 10.4 would prepared for all individuals of school age not be particularly useful for policy makes The attending school, and also for thwse individuals drop-out rate is defined as the number of chil- who have left school in the current school year, dren who left school in the current school year that is, for recent dropouts. divided by the total number of cbildren enrolled 1030 One final educational indicator which can in the current year (plus the dropouts). Again, be computed from the PS is the relation between the tabulations should be prepared separately for age and school grade. In each country, there is an primary and secondary schooling and disaggre- accepted relation between the age of the child gatedby sex and his or her school grade. If children have The PS gives an opportnty for the ana- dropped out of education for periods in the past, lyst to explore why some school-age children are or if they have been obliged to repeat gradep, not currently attending school. For those not in they may find thmsrelves in school grdes which X89

194 Fme 1O.& Age/grade mismathe as a some educational ress So for every hild of prportion of toil eurombents, by SEC, school age and cmurenfy atktndig school, tfis educa lemels and sex' age/gade assemet must be made There ae two broad approahes to tabulating data in order SECI _ SECII Toitd to reveal dfferens in age/grade expriences across socio-economic groups. The first Is to M compute for each SEG the number of children p whose age is above that for their current grade as T a proportion of the total number of enrolled cls4wmxy dren. This is shown in Frame M 1031 Altematively, a matrix can be prepared p T for each soco-econordc group, distributing the sample by age and grade. Such a matrix is illus- Now Thestachastonlfordagedabowtthtd. trated in Frame If there were no agegade slippage in the socio-economin group, a11 the sanple would be across the diagonal But with Frame 1.7: Age-rade matdix for SEG 1 temporary absences and repeats, some chiren will be observed to the south-west of the diago- Primwy edation gvaa nal. Ondy a few children in grades ahead of their age would appear to the north-east of the diago- 1 ; odnal. AmatrixsuchasFramel0.7shouldbepre- AV pared for each SEC. Note, however, that the s relation between grade and age is ikely to be 6 county-specific. Frame 10.7 reports only mis- 7 match for primary education, since this is Ukely a to be a sensitive indicator, and to signal early 10 educational problems for the households con- 11 cemed. Moreover, ndmismatch more wonmon in 12 Ceds comprise nmbersin sample secondaty schooling in much of Africa i The PS questionnaire obtains information 14 for the household as a whole on school-fee exis penditures. This can be used to identify differences across socio-economic groups in expenditureoneducation. Education expenditure in each FIme 10: Mean household education household should be normalized by the number ependiture per cuaently enrolle4 child, by of children of school age in the household. These sodo- oml group and region dali can be presented along the lines of FPam 103. An altemative to Frm 103 would be to SEC ~ 1..- ii Total present the share of education in total expendi- RepioI tures, which some analysts may prefer. Notice 1. that the results should be disaggregated by re- 2. Cells cbdn men ependltn values gion, since it is likely that expenditures will be lugher in uthan areas 4. (espedally in the capital ToW city) where private schooling is more readily available In many countries, policy refolms indude elements of cost recovery in the pubc sector, are inappropriate for their ag. This indicator is and these are often appiied to the education seca futer refnmet of the enolment and drop- tor. To what extent can households afford to pay out indiators since It reveals weaknese i the for the education of their children? How wiling eduction expeiene of those that are cuntly are households to do so? These issue are vital, enrled attending school. For exampe, te since thy determne wheter cost rewvery in erolment and droput rates will no pick up, education wi lead to lower enrolment rates and changes in fte Incidence of repeats, which reflect inureasedropouts among som sodleconontic 'W

195 groups. The information on education expendi- also estblishes the type of tonsultation that is tures wil give govenuments some preliminary made- whether It is firough traditional or modindication of the ypes of houselwid which com- em health services. Health status is a more remit resources to the education of their children sponsive indicator than education performance and the significance of such costs in their overall variables The health of a population can deteriohousehold budgets. The indications are prelimi- rate relatively quiddy, as compared with literacy nary since a more detailed investigation (of the or nuneracy rates, for example. The main probsort proposed in the IS) would be required for a lem with health indicators, however, lies in obthoroughgoing analysis (for example, the PS does taining reasonably accurate information in a surnot obtain information on whether the schooling vey of the PS type. This is because only one of each child is public or private). To account for respondent is selected per household, and it is variations in the numbers of children at school, unlikely that he or she will be familiar with the household expenditures should be divided by health condition of all household members. the number of currently enrolled children in the Moreover, since illness and injury have to be selfhousehold, and the mean values reported in the reported in such household surveys, indicators table. The questionnaire also obtains informa- of health status tend to be relatively unreliable. tion on whether education expenditure has For example, better-educated households tend to changed over the previous year. Appropriate report more illness than poorer and less welltabulations should identify the incidence of in- educated households. The PS therefore does not creases and decreases in expenditure across so- attempt to obtain infornation on the health of the do-economic groups The data obtained in the PS will permit menbers of the household, but concentrates on their recent health consultations. Clearly, some some multivariate analysis, although it should be members will havebeen ill or inured even though emphasized that this is likely to be of a pweimi- they have not had a health consultation during nary nature. Chapter Six reviews the main con- the referece period, and these will be missed in siderations which are applied in specifying an the survey. This focus is also justified on the esimafion model for education analysis, and we grounds that health consultations may be influshal not repeat this here. However, it is worth enced by policy reforms under a structural adnoting the data in the PS which can be utilized for justment progm. Whether these reforms insuh estimation. Our main interest is in identify- volve significant cuts in programs, or the re-strucing the dependent and explanatory variables turing of health services, it is certain that they available from the PS which can be used in Ordi- will change the access of the various socio-econary Least Squares (OLS) estimation. Three de- nomic groups to health care. The PS questionpendent variables are available. First, the pro- naire is therefore designed to pick up these porton of children of school age in each house- changes. hold currently enrolled; second, the proportion 1036 Two broad types of analysis of these data of dropouts in each household; and third, the are required. First, the incidence of health conshare of household expenditure spent on educa- sultations must be computed and compared tion. The occupation, education and income lev- across the SEGs. The problem here lies in interels of the household head and spouse are gener- preting the data, since it is not possible to obtain ally found to be important explanatory variables, information on the incidence of illness. Frame and these are available in the PS. Household 10.9 (overleaf) gives some indication of what is assets (and income) can also be used as explana- required. By incidence is meant simply the numtory variables. Regional dummy vanables may bers of persons who had a health consultation be used to capture the effects of variations in during the past two weeks divided by the total educational services. population in each group. Notice that Frame 10.9 disaggregates by age and sex in order to isolate Helth demographic effects. Three of these are noteworthy. infants and young children are more likely 1035 The focus of the PS is on the use of health to receive medical attention than older children servies by the household, rather than the health and adults; child-bearing women are more likely statusofitsmembers. Infonnation isobtainedon to have medica Dnsultations; and older agehealth consultations of each household member groups will probably nake greater calls on the over the two weeks prior to the survey. The PS medical services 191

196 Fame 10.9: Incdence of health consultations Frame 1011: M per capital household by SEG, age and sex expenditure on health by socecnomic Prou and resion SSCG.oSEGU T 645 Mt 1. F M 3. Ce expenlt.a det value F M.Tol F >60 M F sultationbyeachhousehold membareacorded Total in Secion 1 of the questionnaire. This information wil establish more directly what each tpe of healti Consulstatlon costs the household. rame 10.10: Distibution of health consulta rame provides sonae important intions by type of consultation and soco- frmation about the supply-side of household economic group health epntus lhe cosb of health consul- (PeiW) tations will be expected to vary across types of consultation, so variations down the columns of _EG1 t._ Total _SEG Framte are to be expected. However, var- Traditional heal ations across sodoecnomic groups for each type Health ssaba ofconsultationwillhaveg reatersinficince Midwffe/nrws they wil reflect how accesible the services are in economic terrs Simply because health facilities TOthle1r0 100 are -aalable does not imply that they are access'ble to al households. The table will show whemer health costs are such as to enable poore The second type of analysis permitted by groups to have effectie access to them. Chans the PS concens the type of consultation that is in these costs measured over me with repeated made. This is particularly relevant in situations priority surveys wiql help monitor changes in of wstrucraladjustment,sicethisanalysisshould health costs, and how these chnges are afftng give some indication of the availability of mod- each of the sono-eonomic groups. em heath services to households. Differene in 10A0 PSdata wialsopenitsomeprlinry the type of consultation will reflect inter ala, the multivariate analysis of health. Chapter Five of availability of medical seices. As a monitoring this Volume reviews the types of behavioral rlainstunment, erefore, the PS can signal any no- tionsthatmightbeestimatedforhealth,andtthere ticeable deterioration in the use and (by implica- is little point in repeafting this hem Dependent tion) the availability of modem health services, variabls for the analysis of health uslng the PS Frame outines how this analysis may be data may be defined at the individual and the preseted. househld levels. The individual-leve depend Finaly, health analysis of the PS should entvariableconespondsto thedemandforhealth utilize information on health expenditues by the inputs discussed in Chapter Five, tis being household. There are two estmates of health whether the individual has consulted a health expenditure recorded through the PS question- practitioner during the refeence week. This is a naire. First, as with education, the questionnaire dichotomous variale, taking the value of 0 for obtains information on total houehold medical individuals who have not consulted a practiffoexpenses incurred over the past 10 nonths. Frame ner and 1 for those who have. Multivariateanaly shows how these data should be tabulated. sis therefore requires binary logit estimation And as with education, other tabulations show- methods. Using per capita household health ing the incidence of changes in expenditure (over expenditure (or its share in total expenditure) is the previous year) should also be prepared. Sec- an alternative dependent variable, though this is ond, the costs associated with the last health con- defined at the household leve. As ths is a con- 192

197 Fame 1O.T Averae health costs per consul- Frme 10.T Percetage of ildren (aged 3-60 taton by tpe of consultatio and socio- months) indiatng ignificantly low weight economic group for height, by SEG and riok (pfft) 1b.feau1uIi SEG t_.._...seg 1 ToeA - SETMLf...f...EG..E..Ta.SG I----- SEG 11 Toad Trad_oav heale- Halth assatav Celsetanaeragecb R Midwife/urm 1- Dockwr 2. Other 3 Total 4. Toted tinuous variable., ordinary least squares proce- N t.c euff ziodwsbe75a of saadad lmd*kt A Mfvs att.of may be alelah (for wanmph 8% dures wil generally apply. motwwm dl5 to IOAI The selection of regressors will depend on the level of analysis. With an individual-level dependent variable (.e., consultation with prac- obtain reliable infomkation on illness and injury titioner), explanato variables wiu be of three of each household member, there is one aspect of types: health widch the PS is designed to nmnitor, this * fist, there are the characteistics of the indi- being nutrition. This is one of the "outcome" invidual concened, such as age, sex, relationship dicators that can be readily measured in a oneto household head, work status, etc.; visit survey of the PS type. Anthopometric in- * second, household-level varables should be formation is obined In the PS questionnaire on specfied (indluding education/employment of all children aged 3 to 60 monts. These data household head and spouse, wages/incomes of cover age, height and weight and can be used to household head/pouse and household, land- assess the extent of malnutrition, reflecwt both holding expenditures on food, water sourme used unwngand wastg. Agai, the analytical underby household); pinnngs of data analysis on nutrition are dis- * and third, a set of variables which are exter- cussed in Chapter Seven of this Volume, and nal to the houseold, ideafy induding prices of there is lttle advante in repeat the discusfoods, costs of health consultations and access to sion here. Two broad indicators should be cominfrastuctral services, such as health-care fa- puted fiom these data - height for age and duties. weight for height The former measures the de- Most of these variables ane available in the PS, greeof stunting, and the lattrmeasures wasting. the maior exceptions being some of those that are A child can be conidered severey stunted if his ternalto the household. Healthcostsareavail- or her height is less than 85% of the standard able, but information on inr servies heightfrhisorherage(thestandardbeingdrawn would onlybeavailableif acommundtysurveyis from a reference population - see Chapter alsoconducted. Alternatively,reglonaldummies Seven). Sindmily, a chid can be considered sendght be used as proxies for these influences. verely wasted if his or her weight is less than 75% With a household-level dependent var- of the standard weight forhisor her height Such able (health expenditure), the individual-level children can be considered moderately malnourregressors would be dropped. However, vari- ished. ables reflecting he age/sexstructureof fe house PS data analysis requires that evidence of hold (such as the dependency.atio) might be malnutrition is identlfied for each of the socdoincluded among the houswhol-level regremsors ecodnmcgroups, preferably by regkon, usboth totakeaccountoftmedemographicdetermnants height-for-age and weight-for-height indicators. of health epeture. The household and exr- Of all indicators in the PS, this requires an imnienalindependentvarlables spelfled abovshould diate policy response if there are problems sigbe retated for the household-eve analysis. naled in the indi. The reona d - tion will give more predse dues to policy makers Nutritin about the spedfic grups which are in need of food targeting. Frame reports the sort of 1OA3 Althoug t was stated that the PS cannot tabulation that is required for wasting. The same * ~~~~~~~193

198 ame lo14: Distribution of houstholds by economic sss on the put of the household On house tenure and soenoic group the ot hand, households buying previoudy 4wa2nt) rented accmmdonarelikelytobeinincreasirngly favorable econindc c st How- SEG_l_. SEG_11 ever, the interpretaton of tis table is certain to be comtry-specfc. In some Afhican countries, Tenant rented acdommodation is chosen mainly by the Rtntfree acoatodation wealthy, so that chages toward renting may not SquaWtn indicate economic Wtrs Fmally, the PS obtains infomation on the access of the household to certain basic ameni- F}rame 10.15: Distribution of chd es In ties. Againthesurveyobtainsretrospectivedata tenrialar gementsfor household dweing on the situation 12 months previously. Frames bysocoecomicroup 10.16, 10.17, and report the relevant (FTmO tabulations. Foreachof the aweiutiescovered by -- - the PS, an attempt nust be made to identify Tamwidu SEG.... T.. Tt. O...SEG changes which indicate vements and deteriorations in the living standards of the house- Ito teacy holds. Much of this is county-sc, although From tenany to the Frames illustrate some of the ag- OWfltOCtW gregations that may prove useful. Other dhnges No anes Tolai Employment Employment infornation is obtained in the PS at different levels table should also be computed for stunting. (i) The most impornt information - main eco As with educaton and heath some mlti- nomic activity and job search - is collected for variate analysis should be posswle for nutrition each household meber in the roster. utingthepsddta. Thedependtvaiableinthis (ii) A more detailed descripton of the main job case would be the weight-for-height and height- (including wages), as well as Information on the for-ageindicatorsalreadydlscussed,andtheseof seconday and previous jobs, is obtained for the course are defined at the individual level In head of household and the spouse. general, the set of explanatory variables defined (iii) Infomatonondgrationand total Weearmnfir health atblysis (in paragraphs through ings is collected at the household leve. 1IOAI) should be retand here -these being in One of the major ways though which dividual-level, household-level and etnal cate- stucural adjustment affects households is by gories of regressors. altering their employment opportunities, whch then induces change in the time allocation of Housig *household membes across different types of employment and different sectors of the econ The final basic needs variable that is cov- omy. Such changes will not be the same across ered in the PS cowcerns housing. The PS ques- all households because diferent SEGs have diftionnre obtains infonnation on the tenure ar- ferept endowments, which will influence their rangenents for the dwelling and the utilities that ability to respond to a new incntive structure. are available. Concerning the dwelling itself, the The pattern of edonomic activity across SW4s, PS obtains retrospective infrmation on the ten- and how it changes, is thus one of the key tables ural arrangems 12 months previously. Thus, in assessing the impact of adjusrmt. tabulations on the current situation and the Frame1021laysoutforeahSEGandfor change over the previous 12 months should be the country as a whole the distribution of indiprepared. These are Mustrated ih Fraes vidualsoverbroadlydeflndecotmicactivities and Since the SEGs (the columns of the table) have aearly, any significant change from been defined partly by the main economic activowner-occupation to tenancies indicates some ity of the head of household and since the rows

199 Frame 10.16: DIs(ibudon of households by Frame 1Q1& Distribution of households by source of water nod oco-ecomlc group type of cooking fuel and soclo-econoanic group Water sour * SEG l 1.T Co0kbwjud SEG E..It Tota Riverllake Woedcroal Well Koroes Pubcq top Gas Own tap m'clit.y 0fer O'0 Total Total Frme 10.19: Distribution of dhanges in Fr ame10.17: Distribution of chngslwae sources by socio-econoiic group ( cooking fuel by soceconomic group (-)... Fuddsag SEGIC... _..-SEGl1 1..,,. Totd wau FProm wood/krwe/ somwcha SEG l. 11 Total chmaroaltogau/ electrdty rom river/lake/wel - From gas/electrilty to topubft/privatebtp wood e/ FPrm puc/pvatetop tapl tohr/lalke/well Other changes Frm public t op to Nodhges plivae top Phmrmrivatetopto Total 10D puble tap Nodaunges Total t Frame 10.l2 Distribution of individuals (age 7 and older) by main economic activity and SEG (OMM are acivity categories foraf household members Adii SEG 1..._.SEG U Tota (older than ), the table directly shows the extent to which the household has differentiated its ac- Fa Other slf-employed tivities relative to that of the head. For exanple, Unpaidfamly worker one may find that among the nrral SEGs, food/ Govmentemtoyree - subsistence farm households have very few Ptv esr employee nmebers who do not work as farmers or unpaid Stident family workers, whilst among export-oriented H farmers a greater percentage of household mem- No ecodcac 100 bers have other activities. Such a finding would TOt imply that the latter group has a better chance of witistanding an econondc shock in the agriculturalsectdrand/orhasmoresuccessfullyadapted A useful extension of Frame is to to past shocks. split each row according to gender. There are 1052 In urban areas, the table may reveal, for strong arguments to expect that structral adexanple, in which SEGs formal sector employ- justment affects men and women differently. A menthbas ddnished (a frequent result of adjust- detailed analysis of how and to what extent this ment) and where a growing number of people occurs can only be done with the ls (see Chapter have been able to switch to self-employment. To Eight), but the PS can provide a first insight on the extent that full-time employment is used as a major differences in ecnomic activity patterns row category, the tablemay also give a first indi- between men and womnim, and in which SE(ts cation of low school attendance in certain groups. these differences are most pronounced. 195

200 tam 0: Uenemoyment rat by age, sex ontejob,whohavekotemploynt becuse of- - and SSW.= -EC 1S1G ~ 1037 lhe SDA prototype PS sees more detail SECI the emrplommtpemnof thehead and spouse on 15 ao M F - -presum d to be the main earers In the household ad t ical persons whws employment M must be watdhed for any adverse effects from F adustm t. One effectd ay be that tey cannot 26%30 M find employment at the place of Wesidence and M havq to be away for extended periods of time to p work or to look for work. Such a situation may M be soilly disuptive and have adverse side- F fects on the educaion and health of children. Total M Ai'ays should investigate the Iiente and Total dsbtriution of tis effect through abulr analy- Totia ~~~~~~~~~~~~~~~sis The occupationallassication used in th PS is nwe detl than tme ow that was pos If the PS is used in its mnitoring verion, sible in the househod roster, and the fis anafte chang in the patterns of acivity from one lytic task Is to tabulae the occupational distribuyear to the next can be shown for each SEC, by tion within each SEG, separately for the head of repeating the column of Frame both in houseold and fte spouse. Even though, stricly its basic and gender-diftfeentiated versions- speakin& this tabulation is not repsentativefor for the second year. the population at large, it is pottially very use Since struchtral adjustmnt induces re- ful to obseve the extent to which the resource source switchig, is Important toidentiftypes shifts aimed for by adjustment ane actuay takof individuals and houholds who have failed Ingple, and whetier they take plce within au to make such a switch successuly, and as a r- SEGs. Ideally, thefore, the occupation list suit have become unemployed. This can be should reflectthetadablesnontradablesdistncacheved with the roter questions on lack of tion aemss whh major resource shifts occur work and job sqarch in the ast sevn days (which under adjtmt. together define unemployment according to the 1059 For a wrvey of this type, the occupation ILO defniton). Frame shows a useful dise- list will perhaps distinguish occupations play of this information, abowing the pinpoint- (Ie., a level of disaggregtion correponding ing of the age/sex groups as wel as the SEGs roughly to the 2-digit dassfication of occupawithin which unemployment is a problem. tions in 150,,the bulk of which wiul be non The row totab give the demographic dis- agricudtua occupations. Witin agriculture, the persion of uneploymet, whle the column to- occupations wil be dmilar to the SEC cdassificatals give the indidence of unemployment in each tion, e.g., fammer, pastorlist, and agricultural SEG as well as for the country as a whole. In the worker will be among the occupations,and these monitori PS, this table would be able to show frm also pat of the SEG criteria. Similary, where unemplyment is worsening or improv- within the categozy harmer,i the occupation list ing. A useful refineume of Fiame 1021 is to wil dffenftit acding to major crop, which further split each row (but espedally the rows is also part of the SEG scheme. In other words, a pertang to young adults) according to the level table a the occupation of the head of education achieved. This would the indite of hous d with SEW wil be largey tautlogiwhether the unemploymt pbem is more or cal in rral area& In contrast, In urban areas the less severe for well-educated people (especialy occupation list will provide much refinement recent aduates). This is important information within the SEGs. C tly, different occufor the design of employment and re-taidning patonaltabesforrralandurbanareasarecalled progranms These wil be very different indeed if, for. Frame 1022a shows the dibution of the for example, the majrty of unemployed are re. spouse's main occupation for rural SEGs (which cent tehcl school graduates as opposed to - we remind - were defined on the basis of the people past age 40, with sldlls mainly aediw head'soccupation), hilkatgtheetentto whdch 196-''' "

201 - ae 1Q22a: Main occupation of the spouse rame t0.23 PM"s of oc patona changes in ural SEGs in the last x years, by occupadon SEG I. 6 SEC ToWl o;on * t -~~~~~~2,2 3 ToJg Occupation O0uto2 Pvo occpamtion Total Frame 1.22b: Main accupatio of the head of Totl household and spouse In urban SEGs (Mr"-- SEc SEG 1I TOW Frame 1024: Incidente of occupational changes in the last x years, by cunrent main Unispuu Hed SPMs HOWd spou occupation and SEG Ocaton j ' Occupation Mel axiqutim SEG SEC 11 AM Totad D the spouses work is differ6t from the head's AB main activities. As argued above, diffeed ntiate workpattmrswithin thehousehld providemore protection against chage in the economic environment. An alternative to Frame 10.22a would cross-tabulated against the prvious one; in orbe to crostabulate directly the occtpatiom of the der to be of manage size this table has to be head of household with that of the spouse, al- aed across occupations. The entries are, though thenthefowmalink withthesegclassfi- for each current main occupation, the distribucation is lost. tion of people (ie., heads of household and For urban areas, Frame 1022b shows the spouses) according to their previous occupation main occupation of head of household and spouse -allowing of course for the possibility that there by SEC. Severa altenative versons of this frame was no previous occupation. Job mobility is indiare possible, replacing (or combining) occupa- cated by non-zero off-diagonal cells (the main tion wnth work status or industry, diagonal is empty by defnition). The advantage The PS also seeks data to assess shifts in of this presentation is tiat *t shows th occupaocupation even without having a monitoring PS tional flows. On the other hand, it ladks the link available. SinceItIsaskedhowlongapersonhas to SEGs. For that reason, an additional table is in been engaged in his curent main occupation and order (Frame 10.24) to show, for each SEG, the what the previous occupation was, the two can percentage of people who have been in their curbe combined in a table, using a relevant time cut- rent occupation for less than the chosen cut-off off point of x years ago. The cut-off point can be number of years. This brings out which occupaso chosen that it corresponds to an event relating tions have had recent acrauals and in which SEGs tostructuraladpsent,suchastheiaunchlngof ths ocacred. The "al column shows total rean ecnomic recovery progam If such a period cent accruals for each occupation, and the "all" is too distant in the past, the table could also be row shows average job mobility for each SEC, consted foramoreent period (say, the last ie, the pementage of heads of household and one or two years) to show more recent job tum- spouses who changed jobs in the last x years. A over. The information can be displayed in two complemento this table would be to use previways. ous occupations as rows and show which occu In Frame 1023 the curent occupation is pations have been aboned. 197

202 Frame: 12 Icdence of snday job topic focsing on whether the housewold has holding, by SIG and current main occupation migrated in the last 12 ontws (and if so, where It came from), and/or whether nore rectly, say, C md in the last 3 months, a household memberhasleft e_*jn SEC_ AN the wehold to look for work. Especally the da:upaton 1 latter variable can be seen as a distress vadre in Occaon 2 that, cet paribus, people would prefer to obtain jobs near their place of residence. The ind-.dence of ths variable across SEGs can give an Indication of the extent to which local employment opportunities ae missing for some SEGs. Frame 10.26: Pattern of secondary job holding,.iq65 The tables discussed above are examples by curent main occupation of how the PS infprmation can be used to investi- gate how structural adjustment affects the patcurrantm cat temns of employment. Box 10.2 provides a more SW=dMoyWaiiax AM...*.. _ complete set of tables concerned with ths Issue The employmet section is one of the few I in the PS which lends itself to multivariate analy- 2 ds, in addition to t teab ons suested above. In particular, the fofowing variables - impor- ToWal 10D 100 tant indicators of the employment situation - can be used as depedent variables In simple NcfaTW*tble is caluaed otidy overthae ihvuals wt a models of household behavior. e-odayio * labor fore participation (all membes) a recet occupational change (ead and spouse) * secondary job holding ead and spouse) It has been aued fthat structuril adjust- * wage earnings (head and spouse) nent does not only increase job mobility but also * absence from the household for job search job insecurity, thereby inducing people to hold (head and spouse; household). more seonday jobs PS data pemit the con- With the information available in the PS, each of structlon of tables which display the extent to these variables can be the subject of a simple which this occurs, and in which occupations and multivariate nodel Which may throw moe light SEGs this is concentrated. Fhrme Indicates on the deterininants*f the phenomenon. Several the inddence of secondary job holdings both by of these models are also discussed, in a more SEC and current main occupation. This table can complete way, in the IS-based Analysis Plan on pertjih to the head of household and spoue Employment and Earngs (Chapter Four). The combined, but it is recommended to also con- discussion there on econometric issues lof estistruct separate tables for the head and-for the mating models with binay dependent variables, spouse bece the iniddence of second job hold- aind selectivity bias in the estimation of earnngs ings is lbiwy to be quite different. The presenta- fuictions,is alo relevant here (see also Grootaert tion in Frame is SEC-specific, but it does (1986) for a more detailed discussion). Here we not ident the prevalent primary-secondary job limit ourselves to a brief description of the esticombinations. The latter is achieved by Frame matimg equations based on the variables at hand 1026 which cross-tabulates main. with seconday in the PS. occupation. (To keep the size of this table man Only one model can be estimated for all ageable some aggreiation, particularly of main household nmems, namely, the labor f pr-e occupations, will be necessary.) As with Frame ticipation (LFP). This moel is usually estimated 10.24, we recommend also constructing separate over all household members who are 7 yeaws of tables for head and spouse since prevailng job age or older (this age cutoff can be chraged in combinations wiu not necessaily be the sanme. the light of specific country situations) and who A final element of mobility which could are not full-time students. The latter exdusion is be related to stuctural adjustment is geographi- necessary becawse otherwise the model would cal, Le., migration. In the PS only a very limted confuse labor force participation with the sehoolamount of information can be collected on this ing decision. The dependent variable isbinary: it ia 198

203 .ox :O.2 Tabulation -. the hoehold itself (these providing reedy ac- Box 102: Tabulation plan - employmcnt cess to unpaid family work). These household md orac stis can be entered in the equaton as regressors separately or cmnbined through the * distribution of Indivduals by man ecnnomic activ- SEG lassification whih in fact sunm^m thm * disibuton of ndividuals by nudn emomic acipyacie we recommend that analysts test both tya ndag/sex spedfications to see which gives the most mean- * disdribton of tndivduas by main eoconmi actlv- lngful resuts. ity and educaion level The estinfting equation thus looks as fol- * unemployment rate by age/sex and SBC lows * unemployment re by education level and SEC * average ware Iome and drection of ehane by LFP = f(age, gender, relation to head of house- * ShaeofwagesIntoancmebySEG hold, education; region, household * absen of head of hosehold and spouse by reason enterpsfles) or and SEG * main occpatin of head of houwhod and spouse LFP = f(age, gender, relation to head of housebyssec hold, education; SEG). * wok statss of hmd of household and spoue by * ind ohouehd andspousebysec Except for age, ai the explanatory variables are * eragewag of heed houshooldand of seby discontinuous and thus enter the regression as ocwuation/seg sets of categorical (dummy) variables. * reent oopational changes by curent occupatin Two altematives to this model are feasible * ren occupational changes by SEG (with the same set of explanatory variables). Ih- * cldemofscoaryjobholdangbysegadar- stead of explaining the participation/non-parrent ain occuation ticipation decsion, one may single out a particupatters of seondary job hols by cut main Jar mode of participation and examine its deter- * duron of main occupaton and previous occp minants, for exmple, comparing wage employon of hed of household and pse by SEC mnt to other forns of participation. A second * main ocuaon of headof bmshold axd spouse alternative is to focus on unemployment and by pevious occupation define the dependent variable as unemployed m * remen migrtionbyptwceoforiginand SEG 1, employed =0. This model would provide a * number of household nembers who lft household better explanation of unemployment than was for job search byse possible in Frame 1021, by controlling at the NtAlltablespertainingtotheheadofhousehold sae time for various personal and household and spouse may also be tabuated separately for head characteristics. and spouse, in order to highlight gender differences The next set of models petins to the head of homsehold and the spouse and derives the bulk of the information from the PS questionnaire. The most important dependent variables are occupational change, the holding of a secontakes a value of zero if the roster question on dary job, wage earnings and ebsence from the main activity was answered with "housekeep- household for work resos AAUof these models ing' or "no occupation, and a value of one oth- (except wage earings) will indude the basic erwise. In other words, labor force paricipation demographic variables, altough the relationship is defined to include work for wages, self em- variable is of cwurse limited here to one dumny ployment and unpaid family work. The explana- variable (spouse = 1, head = 0). In the case of tory variables consist of the demographic and wage eanigsn, the best specification is provided education it of the individual (taken by the human captal model, which explains from the roster), as well as selected household wages as a function of education and job experichadacteristics which may influence the decision ence. to participate in the labor market, such as region The dependent variable of the ocupational (employment opportunities may vary signifi- change model is binary and equals one if an occantly across diffeent parts of the country), and cupational change occurmed inte 4 x years, the holdin of farm or non-farm enterprises in with x being defined In the same way as for 199

204 Frames and 1024 This ensures that the absence. The estimating equation is as folows: regssion is corsistent with the tables and provides a furthe explanation. A major determi- Absence a f(age, gender, relation, education; nant of ocuptioral change wil be the qituation occupatbon, work status, industry;occupatloral prior to the change, in particular, the previous change, dgration, change in asses; SEG). occupation, whdch is also recorded in the PS. Anothterrelevant actoriswhether the household A model similar to the one above would has migated. However, the PS only asks for be based on the household-level variable from nigrations that occurred in the past 12 months, the PS, indicating the number of household hence this variable can only be included if the members who have been away for work reasons. model pertains to occupational changes within The differences are that here the dependent varithe past year. Since migration is an easily re- able is continuous and the explanatory variables membered event, it is definitely an option for are limited to the household level. Of specific questionnaire design to extend the recall period, relevancecouldbetheassetsituadonof thehousewhch would permit studying the link between hold, especlally with respect to productive asoccupational change and migration overa longer sets. Asan option the personal and occupational period. Thecstimatingequation is the following: characteristics of the head of household could be included. The basic estimating equations would Occupational change = f(age, gender, relation, be as follows: education, previous occupation; Inigration]). Number of absent household members Whether a person holds a second job de- f(agrlcultural enterprise, number of non-agripends mainly upon his personal characteristics cultural enterprises, assets, change in assets, as wel as upon the nature of the main pob he/she migration; SEG). currently holds (especially the earnings from that job). In addition, the socio-economic status of the Income and assets household can detennine the access to a second job, for example, the place of residence and Income is an important variable in the whether thehouseholdownsanenterprise. These analysis of social welfare, particulay during features are of course captured by the SEGclassl- periods of stnrtural adjusmnt. Not only is flcaion. The second-job model has a binary de- there every prospect of change in real incomes pendent variable, equalling one if a second job is during adjustment, but the sources of income are held and zero otherwise. This model comple- also likely to shift significantly as poliy reforms ments the tabulations in Frames and 10.2 change the structure of incentives in the econ- The estimating equation is: omy. Households previously relying on nontadable sectors fora livelihood will be obliged to Second jb = f(age, gender, relation, educa- switch into tradable sectors as the adjustment tion; current main occupation, duraion, work policies take effect. Yet income is known to be a status, Industry, wage; SEG). difficult variable to obtain through household surveys. It is typically under-reported in most The final model pertaining to the head of surveys for a number of reasons. Respondents household and the spouse serves to explain may not fully know the incomes of all household whether either has been absent from the house- members, they may deliberately understate inhold for more than 3 monfts during the past come for fear of taxation implications, and they year. The binary dependent variable can be con- imply find it difficult to recall income over the structed either to distinguish absence/preserne, spedfied reference period. Income is also subor to focus on absence for work reasons onlry ject to seasonal and other types of variation, so ZTe variables influencing this are the personal that its nmsuremt becomes particularly difficharacteristicsas well as the characteristics of the cult ina PS-type survey, involving as It does only job cuntly held. In addition, the absence can be 2-3 months of fieldwork. motivated by specific recent economic events such _ That income is an important variable duras the loss of a job or of assets. It can be antici- ing structural adjustment and yet is diffult to pated that the soio-economic status of the house- nmere, represents a challenge for the PS. It is hold would also be relevant in detenniting an important for it tobecovered in the survey, but it 200

205 is difficult to obtain reliable estimt The slu- Frame 10.27: Mean shares of income soure by tio to tis dilemmuadopted in the PS is to in- socioderagc gjup clude income, but not to emphasize its precise (wrst) meawrement as such The focus of the PS is on the sources of income and how these may change SEGL- - c_11 TOW in response to adjustnent. There is an attempt to Agi,hr obtain an approximation to income (mainly be- Extcawps cause this may be a useful classifier and explana- Food sops tory variable in other parts of the analysis), but Uvesock this will not be emphasized greatly here. ah" FO&Wb Income data are available at a number of Wageemployment levels in the PS: Pubicsecor * at the most general level, sources of income Foml privae are obtained in the PS questionnaire; incoome sources are obtained in order of importance, and Rents State and privaft some estimate of the amount of income over the Ts past year is obtained for each source; this section Toal 1G 1W ID100 also identifies whether income from each source is greater or less than that of the previous year; * morespecifictreatmentof incomeisobtained analysis of the social dimensions of adjustment. from three sources: Periods of policy reform and austerity requently wage income of household head/spouse; lead to household membs working In multiple gross revenue of agricultural enterprises; jobs (as discussed above), and cropping patterns gross revenue from non-farm enterprises. might also become increasiny diverified. It is The analysis of PS data should begin with particularly important to establish whether the a general assessnent of the main sources of in- diversification of income sources is greater for come of the household, establishing which are some groups than others there evidence, for the majorchangesobserved. This should then be example, that urba-wsed households previously complemented with somewhat more detailed relying on wage employment diversify towards analysis of three principal components of this agricultural activities - such as food production income - wage income of the household head for own wcnption? and spouse, revenue from agricultual enterprises Second, to what extent is the occupational and revenue from non-agricultural enterprises, status of the household head (or possibly spouse FinaUly, an analysis of the asset holdings of the as well) a usefl indicator of the income soure of household, and their recent changes, rounds off the family? In most cases, the soco-econoic this discussion. classifications will use the characteristics of the household head, but this may not be a useful Sourcs of income representation f the household earns a significant proportion of its Income in activities outside Th*ehousehold is permitted to list up toll thosetypiayassdatedwitht o-ec c primay sources of income, and these are divided group. For example, farming households as a according to pre-coded classifications. The re- group may eam a non-trivial proportion of their spondent is asked to give an estimate of the in- total income outside farming- either from noncome from each source. Although the precise farmenterprisesorftomwagelabor. Frame 1027 value of this estimate is not to be taken too seri- will provide a basis for making sone assesment ously, its order of magnitude should be useful of this. Finally, Fratme 1027 will provide clues enough to make some assessment of the impor- about how the poorer soio-economic groups are tance of various income sources. Frame likely to be affected by policy reforms. If such should be computed from the PS data, indicating groups rely entirely on income fron food crops, the relative importance of each of the income and if the latter are nontraded, it is likely that sourcs in the incomes of each socio-economic such groups wilsuffer during periods of adjustgoup. Frame 1027 will assess three major is- nenl sues. First, to what extent do households diver The PS prototype questinnaire also obsify their income sources? This is critical for the tains some useful information on whehr the 201

206 Fame 10M Poportion of bousehold exper- Frame 1029: Mean Incomechg scomes by - -echigdmages in Ioe by hwome sou aodo-e4ouomlc group ad socinoc group SEC 2 - SEC n T0 Unwiwhtedseres Weighed sers! wmeaoutea (.zaieae gmb", -meowice) Agrclbse Expori crops Inraeese FoDase make sese of these responses in terms of the h, economy-wide dhanges that have occurred dur- Drease ing the recent past. Some impression may be Avestck gained from the table about which of the soidoktae- economic groups apper to be losing out and DUCw whidch (if any) seem to be gaining as a rmsult of Olhap - adjustment. This, however, maybebetter D through the second type of analysis. Ths involves.. Non-farm enterprs compuftng scores, in which +1 indicates aga in Dba_ ~income-and - a loss (no ne isindicated by 0). These scoreshould be aggegat for eachhouse- Wa:s employment hlid and mean scores derived for each sodo- Publiceockw ecnomic group. Ts would give some indication of the chne in overall inoomes being expe- Derem 9rienced by the socio-enonoic groups over the incuw rivea past year (see Frame 1029). Arefinement would -ue be to weight the score with the shaes of the ncome in total household iro=m Thus, a -1 Ina ~ ~ would anmt moe for a prindpal source of in- -S~OUS come than a secondary source. knevas Wage income iiecrq,_ Infomation on the wage eanings of both state ;a9 t-he household head and spouse is reported in the _eaem PS questionnaire. Income from own-account work is not covered in this scion - ordy from Private se - -- wage employment. The questionnaire petmis a.n -e te division of wage income between state and pi- ' t -vate _ sectors. Ths information should be abulated along the lines offrame Tis repors the average wage that is earned by the household income fom each soure has inceased or de- head and the spouse from pubic and privae creased since the previous year. The analysis of sources, the latter by industry. The secral orithis information can follow two tak The first gin of wage inome can be derived from a involves the separate analysis of each souce-of pate quesons in e PS quesioa Inb Income. Thus, for each sodo-ewnomic group, an puting the wage, care should be taken to ensre assessmet is made of the direction of change in ftat the dthe period used is constant across all income by source, as shown in Frame observations. Re-computations! are usually re With inpome from some sources expected quired to recast the wage data into a common to fall and from oths to rise during the struc- tme unit. he data of ame 1030 can be used to tua dmanges brought about by policy reform, assess whether labor-market signals are in line we would expect to find some paens ememng with te main objectives of the structural adjustfrom Frame An attempt should be made to ment prgram In the very short run, wage dif- 292~'

207 FMe UL:- Prpotidon of farms epeuc- u clhaies In prodution and areplapted W Frae 10 0 Men wage Income of household head an4 spowse, by SEG and sector -, - a~~~~~cop SEG l... SEG 11 ToW - liouseholdhead PFuva ase I 2 Inczois DcVase Inave Daaus Maize State Yam Totil lanin Spew. Cels cmltn valuesc Pdvat secdw Uvestock" not appicable 2 * ChanIesn mber fduveswoo Om 12 s pulor to toh y Totailtion. It must be noted that variables related to occupation should not be included for this purpose ferences can be expected, encouraging labor to as they rpresent factors subject to choice by move into tradable-oriented employment. Thus, household mebe. For actual estmaion, the private sectr wages shud (cteris panbus) be dependentvariable (wages)needstobeconvertec higher than state sector wagese and witfin the to a comon time unit and is best expressed in private sector, wages in exportables and import- natual loathm fim: ables should be exped to gain relative to those in nontradables. The table also identifies wage Ln(wages) = f(educatlon, eien, experivariations across the socl-econemic groups. enceqard; gender, region). Clearly, these differences will depend on the characterwsi- s of the workers conerned - par- Agrkulturl activitis ticuarly their human capital Thus, for example, wages may be higher In one SEC than another More detailed analysis of the agricultual smplybeusewageeaersarebetereducated. activities of the household is possible through Tabulations of this sort can only be indicative of the responses to the PS quiaie These spicheffects. heapplicationof multivariatetech- cover production of the main food and export iques, however, may unravel some of the rys- crops, changes in this production and the gross tezy. sales. The section also ascertains whether the The estimation of a simple wage-earnings area planted to each of thecrops has increased or model, based on the human capital theoy, is deceased and why this change has occurred. useful to assess whether the relative returns to -Periods of structural adjusntent are usualy chareducation and work expience have changed acterized by rapid changes in relative pries and during an adjustment period. In the PS, total in the incentive stuctr Whether farmes rework exeience is not available, but the experi- spord to these changes has been a subject of ence in the cunent occupation is known. As this much debate. lhe PS can provide some pelidiis the most relevant in determning current earn- nary evidence of recent changes in production. ings,itisasuitablevariableforestimation. Many Again, in a survey of this kind, the analytcal labor markets in developing countries are seg- focus should be on the change in the levels of m enreil workers cannot easily move between production and areas planted to each crop. This difnt section of the market (eg., between the omationis nmmrized in Fame The hinmal and formal sectoo, between regions) table shows the proportion of each of the crop and/ordifferent typesof workersarenotconsid- producers which experience changes in producered substitutes for one another by employers tion and area planted (the denominator b6i the (e.g., male and female.workers). The human total number of households produing the crop capital mode abows the indusionlof such vari- in queston - that is, the denominator will be ables to test for the presence of such segmenta- different for each row). Frame 1031 wil givefur- 203

208 ther dues about the income ctanges observed in PFame Reasons for change in area Frames and 10,29, at least for agricultural planted by direction of change and aop households. It will show which crops are becon- (6ent) ing less or more profitable. It is also proposed that livestock be included in this table, though here the measure relates to the change in the RCitai Iv Da e osa h*crse Dease stock of livestock over the past 12 months The respondent is also asked for reasons Precagse why the area planted has changed since the last Fe5avaL!ab«q season. From this, some assessnent can be made labor avaulty about the underlying factors behind the changes Mpkedng in crop production and area planted. This infor- creditaviwuty mation ought to be tabulated along the lines of Other Fame 1O32. Of the households reporting changes Totwl increases/ in each crop area planted, the proportions giving dea_cre each of the stipulated reasons are computed. In this way, the table provides some indication of Frame 10.33: roportion of farming housethe predominant factors leading to changes in hl using idpus by type of inputand main crop area (and by implication, production). These crop produced reasons given (which are precoded in the ques- (parrt) tionnaire) will to some extent be country-specific, but itisdearthat some will relate to price changes, and others to the libealization of markets and si' rop Ftgizwr CTediI SesUiLAW the improved access to productive factor. These Maize factors are likely to be particularly subject to Yaii change as a result of structural adjustment. For example, an adjustnent program which liberal- Coffee izes Imports may lead to an improwement in the F a felize lat season. availability of fertilizers, and a consequent in- Credit fmrshavinedat faorming cre in the rate of return to fertilizer-responsive durgi*seasu. crops. Alternatively, it may reduce subsidies on Exnon: s ames visited by ensio worker durng lost agrcultural inputs, which reluce rate i of return. muon Thus area planted may increase or decrease dur- Labor faufm hwing m workes Int s n tha ing adjustment, depending on the induced sea bdow changes in output prices and on the availability (and prices) of inputs The PS questionnaire also offers the op- extenson services, the directit of change deportunity for a more detailed analysis of the pending on the details of the adjustment packchanges experiened by farmers in the use of age. It can also tighten (or slacken) the labor inputs-such as fertliizer, credit, extension serv- market, thus changing the avaiability (and profices and bbor. It is possible to compile responses itability) of labor services. Farmers employing to these questions by principal crop. In this way, more labor than previously are likely to be bene- Frame 1033 tabulates the use of these inputs by fiting from recent policy changs and exanding main crop produced. Farmers are smply asked their operations. These inputs are all brought whether they purchased fertlizer, obtained credit togeffier in Frame or received an extension visit during the previous planting seasonl They are also questioned Non-fl rm enfep,s about the use of labor on thte fam and its change over fte past season. Their responses give some The PS design also permits an assessment indication of how readily available and how prof- of changes occuring in non-farm enterprises (the itable suchinputsare. We have already reviewed curent prototype questionnaireprovides for up how adjtent can improve the access of fam- to three such chunges to be enurated for each ers to fertilizers (and also change their price). But household). As with agricultural activities, non- It can also change the availability of credit and farnenteqrpsesarecertantofacechangingcon- 204

209 straints and opportunities as a result of the ad- Fnme.34i Number of non-farm enteprises justment effort. Market oppoftunities will open by sectr and SEG up for tradables, and decline for nontradables. Ifrastructural changes may improve we tod50 SEG.1 SE G TotlW product, factor and credit makets. These will combine to change the profitability of non-farm 2 cnain tow activities, so that some will expand and others 3 mumber of enterprises identified decline. A critical first stage of the analysis is the - classification of these activities into exportables, ToWl importables and nontradables, since this fonns a vital link back to adjustment policy reforms. A code is recorded for each enterprise which should Frame 135: Proporon of enterprises experireflect these distinctions These codes are inevi- ending output and employment changes by tably country-specific. The next task is purely sector descriptive - to identify which groups operate enterprises in each of the sectors. Frame E shows the type of tabulation that is useful. Its Sea I'e Dwwa Nod b Decrc Nod8g purpose is to identify which SEGs are most in volved In household-based non-farm enterprise, I and in which sectorsare they active. Some judge. 2 ment can be made from the table about which 3 groups are likely to gain from adjustment- these - groups being those that are particularly active in setors likely to gain as a result of policy reforms. to be included in the list are of course country Whether or not an activity benefits from specific and should be judiciouslychosensoasto adjustment can be assessed from the PS findings represe.t the best possibl- proxies for total housethemselves. The questionnaire identifies which hold welfare. Since the IS contains a more conenterprises have increased the value of the gross plete list of assets and also requests espondents output during the past 12 months and which to provide a valuation, IS results will permit the have increased the numbers employed. These refiningof theassetlistforlaterprioiity surveys, areusefulindicatorsof whichsectorsareexpand- by calculating correlations between specific asing as a result of the adjustment reforms and are sets or sets of assets and total household income ilustrated in Frame The interpretation of or expenditures. the tabulation is self-evident - sectors in which Expenienoehashownthatvaluationquesoutput and/or emnployment is expanding are tions are difficult and time consundng and are iely tobebenefiting frompolicyreforms. These not an appropriate type of question for the PS. findings should be cross-checked against expec- Moreover, value information really needs to be tations arising from macro-meso analysis. The combined with information on the age of the combind evidence of Frames and 1035 assetinordertocalculate depriation. ltelatter will enable policy makers to make some judge- is an analytic procedure which does not fit with ment about how policy reforms have affected the the rest of the PS analysis. For those reasons, the different sodonomic groups, at least as far as PS questions on assets are linited to two very their non-farm enterprises are concemed. simple questions: does the household own the asset, and has there been any increase or de- Asses crease in ownership over the past year. The second question is meant to provide a direct indica Assets occupy a special place in the PS tor of change in welfare. In the case of deeases, because they provide the only direct measure of this can be an alarm signal that the household is economic welfare and of change in it. For reasons forced to dispose of its assets in order to meet of time and sensitivity no attempt is made in the current consumption needs - a situation which PS to obtain an exhaustive list of household as- is dangerous for the household's fture incomesets. Instead, the list is limited to selected pro- earing abilities. ductive assets (land, farm equipment, etc.) and Any one asset is not likely to be a sufficonsumer durables (car, bicycle, etc.. The items ciently reliable proxy for welfare or change of 205

210 welfare, butu a buyorsell tend isobserved over Frame 10. Asset ownership several assets, the interpretation is strengtheed. For that reason we do not propose tabulating re- SEG I. 1SEG sults by asset, but rather calculating a simple wop sowre consisting of the number of cat6gories of sa ofasse o.me assets owned by the household (proxy fo wel- PrcntagOf houzokb fare level) and a net change score, i.e., the num- whoownland ber of categories with an increase minus the Peass fhousds number of categories with a decrease (proxy for wh own catte welfare changes). Two exceptions to this treat- awngeinownersmp ment can be made for land and cattle, which in Net chage soe of rural settings are of greater importance than other assets owned assets, and which therefore merit being shown P stage of housolds separately. The asset score will range from zero Fercetg w-olos-- la of htousehold (no assets owned) to the number of items in the who gainwe td list (all assets owned). In the case of this proto- Perentage of houshds type questionnaire the score's maximum value is who lost cle 10 (there are II items listed in the questionnaire Percentageof houshold but land is not included in the score calculation). who ganed cate The net change score in principle can range from -10(a decrease in all assets) to +10 (an increase In all assets) but in practice it will vary in a nar- his analysis. The presentation therefore has rerower range around zero. lied to a Lare extent on the use of selected cros tabulations. In ondy a few instances (for example, Conclusion in the case of employment, wages and education) has there been any attempt to suggest multivari TheprioritysurveyoftheSDAisdesigned ate analysis. This is quite deliberate, but does not to obtain priority infonnation as expeditiously as imply that no further multivariate and monre depossible. This has obvious implications for the tailed analysis is possible (and desirable) with Analysis Plan presented above. First, the range the data set generated by the PS. Rather, what of data that is obtained is certain to be somewhat has been suggested here should be Interpreted as resticted, and the analysis has inevitably to rely a basic agenda for PS analysis. It is hoped that more on the use of proxies than in other Analysis this Chapter will prove to be fruitful in country Plans. Second, the critical need to produce policy applications and will stimate extensionsbeyond research documents quickly restricts the time that the cross-tabulations and limited multivariate the analyst has to prepare the data and undertake analysis suggested here. 206

211 Technical appendix: Concepts and estimation of household incomes and expenditures within a system of household accounts ntroduction counts can be constucted to include the revenues and costs of houseold prducion activity A.01 Most analyses concering the impact of (te production account), household incomes, adjustment on the living standards of individu- consumption expenditures and transfers (the als in society utilize, or rely on, measures of current account) and savings and the acquisition household incomes and expenditures. The con- of real and finanial assets (the capital account). ceptual basis for these measurements is broad Records of household assets and habilities may and far-reaching (Grootaert, 1983), not least be- be inworpoated as an extension. The obvious cause the terms income" and "expenditure can attractions of expressing these variables in an have nmrkedly different meaings in different accounting fiamework are that it imposes a concontext In some instances it might simply in- ceptual discipline, it makes the relationships bevolve relating analysis to quite specific compo- tween thecomponents explct and it enables ernents such as "expenditure on food" or 'wage ror assessment to be based on the size of the income," but in others t might require the identi- inevitable disvepancies. fication of much broader measures of "totar in- A.03 A number of issues need to be addressed come or expenditure, perhaps stretching as far as in the calculation of household accounts from the full income, where even leisure is assigned a Integrated Survey (IS). First and foremost it is monetary value (Kusnic and da Vanzo, 1980). important to have dear conceptual definitions of A.02 The SDA survey instruments, and in par- the components of the accounts, a task which is ticular the Integrated Survey, contain a poten- far less straightforward than it appears at first tiauy rich source of information relating to the sight. These conceptual issues are discussed in surement of both monetary and non-mone- detail in paragraphs A04 throuugh A35. Secondly, tary variables. There is, however, a clear need to it is necessary to fit the available infornation as distil and organize the complex responses that closely as possible to these idealized concepts. have to do with the monetary variables before Thus It is necessary to tranm and aggreate analysis can proceed. In view of the wide range the variables in the questonnaire to build up of intepretations and the different needs for estinmtes of thse concepts, which should also subsequent analysis, there is a strong case for correspond dosely with concepts required for organzing these data within the context of a co- purposes of analysis. Nevertheless,itisdeslrable herent yet flexble framework. The most appro- to incorporate sufficient flexibility to enable priate framework relating to the aggregates of slightly different measures to be calculated if rehouseholdincomesandexpendituresisprovided quired. Paragraphs A36 through A.96 discuss by a set of household accounts. In this Appendix the major issues and difficulties involved in proit will be demonstrated how the household ac- ducing these estinutes from the l!questionnaire. 207

212 A detailed piposal for estimting current ac- unltcoudbetooheteonousinitscomposition, count aggregates is set out in the Annex and may not in any sense be a single economic and social unit. Furthermre, the degree of intra- Household accounts Conceptual basis household iequality (of all kinds of variables) could be too great. So a furter criterion is added The household unit for survey and analytical purposes, wich is that au persons living and eating together should ac- A.04 Individuals can, and usually do, belong to knowledge the authority of a single head of several overlapping networks of social units, at household, regdless of whether the latter is the same time (World Bank, 1990: 98100). Nu- living with the other household nembers or livclear and extended families are two such units ing away. Hence, servants and lodgers may or and the household is another. Fromn the point of may not belong to a household depending on view of micro-level inquiry, commonly utilized whether or not they accept the authority of a criteria for defining a household unit include: household head; if not, they should be treated members who have a common source of major and listed asa separate household. In the African income, live under the same roof or within the context the situation of polygamous households same compound and have a common provision canpresentproblemsdependingonwhethereach for other essentials of living (Casley and Lury, wife is treated as a separate household oras part 1987; United Nations NHSCP, 1989). In practice, of one large household. There is no easy answer the precise set of criteria used to identify house- to many of the definitional problems and in holds is chosen to suit the local situation so that practice, the definition is subject to a degree of the size and characteristics of households can arbitrariness, at the margin, not least because show wide variation by locality and country. household mmbrsip is dynamic and there- However, a basic distinction can be drawn be- fore could vary during the time interval to which tween the household and the family unit, where a survey or measurements relate. I for the latter there would need to be an additional A.06 The household is an important social and criterion of kinship. Households could consist of economic unit because: a single family, but commonly In Africa they (i) within it many of the decisions concerning consist of several families, whether they are of individual members' activities and their conthe same kin or have no kin relationship. Also, it sumption (and hence their welfare) are made, is possible for families to be spread between and households either temporarily or pennanently. (ii) its physical properties - the fact that it is a For example, a married woman while young may collection of individuals with an identifiable locontinue to live in her fathers household, while cation - make it a useful sample unit in survey the husband lives under a separate roof; or a son work. may work in the mines or an urban area and The first of these reasons, the decision-making return to the family home for only short periods role, is most important from the conceptual each year. standpoint. The criteria for defining household A.05 The basic United Nations definition of a menbership are broadly sufficiento ensure its household, used in many surveys, is a "group of role as a social unit, although a household might people who live and eat together." While this is well contain a hierarchy of subunits and could clear as regards the distinction between house- inter-penetrate other households through the exholdsand families, there are significant problems tended family. However, the household may not conceming individuals who just board (eat with always coincide with a single economic unit. In the household) or ordy lodge (live with the this regard we can disinguish two kinds of ecohousehold). The United Nations guidelines sug- nonic unit, one relating to the activity of congest thata boarder who does not lodge should be sumption and the other to production. A third included but the lodger who does not board unit relating to the accumulation and ownership should beexcluded, suggestingthatthe"common of assets could also be cdnsidered if it differs pordominatesthe"commonroof'forhousehold from either of the other two. The household membership purposes. A further consequence is definition modified and adopted above prethat those who board and lodge, such as domes- dominantly recognizes the househows function tic servants, should be regarded as household as a consuming unit and, to a lesser extent, as an members. For SDA purpose, such a household asset-owning unit. However, living and eating 208

213 together might not imply the pooling of all in- rated business activity. herefore tis exemplicomes and the shating of anl expenditures. Even fles the earlier point that the appropriate concepif it does, and the houwsehold constitutes a family tual and statistical unit of account from fte point unit, two possible concptsof household dedsion- of view of production may not coincide with the making based on the "glued-tgther" or "des- household as a financng and consuming unit. potic" family unit could lead to widely different There are also inevitable problems with taxonintra-household allocations of resources and in- omy and the dassification of institutional units at dividual well-being. Furthermore, the major the macro evel which serve to emphasize the household assets are usually formally owned by need to define household" units sufficiently a single household member or a sma1l subset of broadly to include all persons who are inmates of the total household membership. Institution such as prisons, asylums or hospitals A.07 The problems stemming from the on a long-tern basis (comunon roof), who share household's role asa production unit can be even their meals (common pot) and respond toa commore problematic and of a different Idnd. This is mon authority. largely due to the fact that those production units usually considered as household-level activities Prtduction boundary and ousehosld (small-scale agricultural holdings or non-farm enterprises) may not match the household unit. A.09 Measures of household incomes and ex- There is no problemin those Instances where the penditures can be approached from different production activity is wholly operated by a sub- perspectives, but traditionally the most tractable set of household members and engages the assets starting point is to consider household activity in of labor and capital owned by that household relation to the boundary of production. This is unit. Indeed it may not even be managed by the the demarcation line between those activities redesignated household head or decision-maker. garded as being productive and those which are But there are problems if a production unit is non-productive. Some might argue that the disoperated by more than one household unit and if tinction is arbitry, but it does have a direct there is any difficulty in identifying and appor- bearing on output and income nmeasurement and tioning the incomes generated. In some instances, our ability to define aggregates at the micro level difficulties on the production side have led in- and,moreover,ontherelationshipofthesemmico vestigators to prefer a broader institutional unit aggregates to (and consistency with) the aggre than the household for survey purposes - for gates for households as a whole at the macro or example, a group of individuals living in a hut in economy-wide level. Various formal defnitions the village or cattle post (Botswana), and a of "production" exist. Hicks (1971) defined it as homestead (Swaziland). In general, however, vany activity directed to the satisfaction of other there is no reason to expect the consumption and people's wants through exchange, a definition production urdts to match which has been a fundamental tenet of national A.08 At the macro level the national economy income methodology for over fift years. The consistsofanumberof institutional sectors These strict interpretation of this definition would exare designated classes of transactor or 'resident clude a whole range of own-acunt activities on economic agent" and are usually grouped for the grounds thattheyarenot fornallyexchanged. national accounts and analytical purposes ac- A.10 However,"exchangehasnotalwaysbeen cording to their similarity of function. In the interpreted in the strict sense, and productive United Nations System of National Accounts activities do indude many of those which are not (SNA)(UNSO, 1968) two main classes of trans- marketed. Indeed, more recently, Hill (1979) has actorare disnguished. Theflrstclassconsistsof suggested a very broad definition whereby an prducingunitswhmethesecondclassdealswith activity is considered as being productive if it financing units, that is, units in which financing could be carried out with comparable results by decisions are taken, relating to both current and an economic unit other than that which acualy capital outlays. Fnancing units include house- performs it. Hence, eating and leeping would holds as a group and are distnct from corpora- be non-productive, but many other personal and tions (defined according to their legal status, usu- household activities, such as washing and cookally relating to corporate ownersip of assets) ing, would be considered productive regardless and government. The definitional classification of the quality of the result achieved (Chadeau of production units usually includes uninpcpo- and Roy, 1986). 209

214 A.11 The definition of the production bound- durables lie on the borderline between being ary adopted from the national accounts stand- goods or services and, indeed, on the production point usually falls far short of the broader defini- boundary itself. Decoration and minor repairs tion suggested by Hill. Marketed goods and (do-it-yourself" activities) are considered proservices are always included, but quite difficult ductive if they are carried out by an owner of a questions surround the extent to which non- dwelling to maintain or add to its value, but not marketed goods and services should be included, otherwise. In a strict sense serices can affect as well as the decision on the appropriate impu- either people or goods and are usually associated tations that should be attached. with a change in condition brought about by the A.12 Goods are different from services. Goods activity of some other economic unit (Hill, 1977). are physical or tangible objects for which a de- Hence, the essence of a service is that one ecomand exists, which are appropriable and there- nomic unit "serves" another. There are no esfore are transferable between persons or other tablished principles on the general treatment of economic units (Hill, 1977). In principle all goods non-marketed services and, unlike non-marketed produced should be included within the produc- goods, they do not tend to arise in the context of tion boundary, although in practice there are sig- imputations at the household level. One sigrdfinificant difficulties in actually doing so. The cant exception is the concept of the production of problems are almost exclusively confined to dwelling services for owner-occupiers, whereby household-level activities where certain informal a service is considered to be both produced and activities related to own-account agricultural, consumed within thehousehold unit. Asno cash pnmary or manufactured goods are not often transaction takes place, in common with other captured. The main argument for including all non-marketed products, the main problem is to goods in this way is that at the time of production derive an imputation for this output. But the it is not always known whether the goods are inclusion of dwelling services does differ signifidestined for the market or home use, these two cantly from other household services in that it is uses being dose substitutes in general. Hence, in attributable to the household by virtue of the a strict sense, the following activities, whether household's status as an owner of a productive marketed or not, would all count as being pro- asset rather than as a household activity per se ductive: (Chadeau and Roy, 1986). (i) production of agricultural goods: including A 15 One other category of non-marketed servstorage of crops; gathering of uncultivated crops; ices that affects households is the provision of forestry, wood-cutting and collection of firewood; collective services (public education and health) hunting and fishing; etc and pure public services (public adminisration (ii) production of other primary goods: mining and defense). Although these are not produced salt; cutting peat; carrying of water; etc. within the household unit, incomes generated in (ii) processing of agricultural goods: production their production are directly attributable to and preservation of meat and fish products, households according to the lbor services they crushing oil seeds, weaving baskets; etc. provide, in just the same way as they do for the (iv) other processing: weaving cloth; making production of all other goods and marketed servfootwear; maidng funiture; etc. ices. But it is usual to record outlays on such A.13 However, in practice, most countries limit sevices as government expenditures even though the scope of estimatng the output of non-mar- they materially affect the well-being of houseketed goods to the activity of farmtrs producing hold members. agricultural products which they consume, val- A.16 The main point to note in the present conued at farm gate prices (Ruggles and Ruggles, text is the extent to which national accounts con- 1986). This is for the simple reason of practical ventions that underlie the concept of production, expediency, and which servesasa first indication the production boundary and, hence, of income of the anomalies in estimating incomes and ex- and expenditure, depend so crucially upon the penditures at the micro or macro level. Some notion of the household unit. At least in prin- African countries do not even include these lim- ciple, virtually all goods produced on own acited imputations in their national accounts. count (non-marketed) can be included within the A.14 The definition of productive services has boundary of production on thegrounds that they proved to be far more debatable. Some activities, could have been produced by some other housesuch as maintenance and repair of dwellings and hlwd. The same is not necessarily true in regard 210

215 td the distineion between produced services and of institutions, or groups of t ns, chosen the ange of other activities performed by house- on the basis of their economic fudcons, legal hold members within the household unit. The status or other characterstc Each insttution convention here-and itisbasicalyya convention can be viewed as an economic system, which -is that the kinds of activities perfomred for one interacts with other Institutions both within and household nmeber by another (but still within outside of the national economy and its boundathe household unit) such as cooking, cleaning, ties. Taken to the limit each household can be teaching childrent, etc,, are not considered as be- viewed as a minuscule economy engaging in the ing productive. Had the 'individual" been se- economic activities of production, consumption lected as the economic unit instead of the and accumulation to either a greater or a lesser 'household," then, according to the strict con- extent, and interacting with the rest of the nation ventions, the situation would be different (Hill, (and the rest of the world) through its trade in 1979; Hawrylyshyn, 1977). Only at the individual conmmodities and factor services and its accumulevel does it become possible to set a precise lationofassetsandliabilities,bothrealandfinanboundary. Non-service activities such as eating, cial. If a household does not engage in housedrinking and sleeping simply cannot be under- hold production, then all of its income must be taken by one individual on behalf of another, derived fron outside in the form of factor income, even though many other activities can. Hence, in respect of, say, the supply of labor services its by taldng the household as the econotnic unit, the members provide, and of non-factor or tansfer boundary is drawn to exclude many individual income. Similarly, itsexpenditureout of incomes level activities from the scope of imputation, and received would be on goods and services prothis has a direct bearing on the types of measures duced outside the household. On the other hand, that can be adopted at both the micro and macro if itdoesengageinproductionof marketedgoods levels of analysis. and services, then the incomes generated inside A.17 This is not, however, to say that service the household would be derived from the disactivities conducted within the household by one posal of the product to users outside. Production member for others (such as housework) are on own account would be internal to the housewordtlessonthecontray,thereisanopportunity hold. Usually no actual transaction takes place cost associated with these activities. One means between household members, and the implicit of recognizing this would be to constmct time use accounts for each Individual, that is to deterincome and expenditure that arises has to be im- puted. mine and, hence, value the pattern of time use for A.19 In an entirely analogous way to the staneach individual in society (Becker, 1965; Pyatt, dard accounting constructs for the nation as a 1990). In this wayactivitiescouldbedistinguished whole, it is conceptually possible and, indeed, between the production of tradable goods and with the appropriate survey capability, feasible, services and nontradables (such as leisure and to assemble a set of economic accounts at the other personal activities). All activities have an household level The discipline one derives from associated price, tradables being valued at their an accounting framework helps in several key market prices and nontradables valued at their respects with the problems encountered in opportunity cost; the latter will differ between choosing the appropriate aggregates of incomes individuals. This is probably an area of work for andexpenditures. Buttheaccounting framework the future, but is highly relevant because it ex- does not and cannot alleviate all the conceptual tends to the very limit the notions underlying the difficulties, some of which arebeyond immediate concept of production. Furthermore, accounts of resolution and are stil the subject of considerable time use and estimates of the associated opportu- debate. The basic accounting structure is the nity costs are relevant to determining how assembly of separate accounts for each of the household members are likely to respond to the economic activities of production, consumption changing economic environment under ad,ust- (income and outlay accounts) and accumulation ment (capital transactions accounts). The principle focus for much of the SDA analysis i on particular The housdeld economy aggregates to do with household consumption and income, and henceon estimates derived from A.18 At the macro level the nation can be subdi- the income and outlay acounts, but it is clear vided into mutually exclusive and exhaustive sets that the interdependence between the three 211

216 principal accounts means that the way in which (labor and capital including land) and supplied the elements are defined and estmated depend both to own-account production and to produccrucially upon what goes into theother accounts tion activities of other nstitutions. As already of thecomplete system Someof thecompnens indicated, factor incomes may be generated In of the other accounts are also of dt interest in two particular ways. First, houseold members their own right. sel tleir labor sevices to production units and A.20 The production account for an Individual receive wages or salaries in return either in cash household should set out the operating balance or in kind. Secordly, they may receive income sheet for all activities deemed to be productive, ftom productive assets which they own. In prac- In principle this should involve a valuation of the tice a major source of this income is rent frwn the output of non-mrketed goods and services as ownership of dwellings; either actual rent or, in well as marketed output. The incomes derived the case of owner-occupie dwelungs, imputed from each activity would then result from the rent. Icome generated by the ousehold's own difference between gross output and the total production activity, referred to earlier, has to be input costs incurred, inluding an allowance for induded in factorincome odmponents but, being thedepreciation of capital assesused in produc- in general a return to both labor and capital, it tion. Household-level production activity is nec- does notcomfortably fitinto either and so maybe essarily unincorporated and is often informal. added as a separate, third category. Some labor is hired, and wage income is paid A.23 Non-factor, or trafer, income can arise (either to household members or outside), but as transers between households or as a receipt frequently it involves remuneration to a member from other institutions such as govenment (eg., of a household who is either an employer or educational schoblarips). Such income receipts own-account worker, in which case the income maybe highly tansient and it maybe difficult to reprsents the return to both labor and capital distinguish them from capital trnsfers at the employed. This constitutes at least part of the margin (eg., gifts received, dowry). So just as household's income, therefore it is represented there are boundary problems In defining the exas an outlay from the production account and as tent of production, there are significant problems a receipt by the household current account. In in defining the boundary between curent and arriving at the estimates the accounting disci- capital items in a household's account. pine means that au Inputs and their associated A.24 OA the outlay side of the household's curcosts have to be properly identified. In many rent account, in addition to cash expenditures by cases this is not problematic, but there are a sig- household members, there will appear "in-kind" nificant number of instnces in practice where it expenditures corresponding to "in-kind" income. is difficult to distinguish (even on conceptual Cearly, this is especially imnportant in respect of grounds) between inputs into household pro- subsistence agriculture, but it is concptualy no duction and the consumption outlays by house- different for home conmption of non-farm hold menmber Quite simply, this is often be- production. Likewise, non-cash incoms in kind cause of the physical congruence of the dwelling (e.g., housing subsidies) should be inluded as and production units, and because separate ac- imputed expenditure. Besides expenditures on counts are rarely assembled for them. goods and services, a household may make cer- A.21 In principle, the household income and tain transfer payments to other households or to ouday account is distinguishable from the pro- other institutions (e.g., direct taxes). Adding toduction account at the level of the individual gther all such outlays, the balce between this household (the micro level) in much the same and total incomne would yield gross savings for way as it is distinguished for any institutional that household, which would simultaneously group at the macro level. Incomes are received appear on the households capital account. from various sources, including from the pro- A25 The capital account for a household comductionaccounttofinanceexpendituresand other pletes the description of its economic activity. outlays, and this account sets out the balance Tere are various possible arrangements, and it sheet for the individual household. These incnmes harbors a number of difficult conceptual probfall into the two basic categodries of factor and lems to do with valuation and inclusion. Essentransfer income. tially, however, the capital acdunt serves to bal- A.22 Factor inoomes are derived from the use ance household savings against the net acquisiof the productive assets owned by the household tion of real and financal assets. National ac- 212

217 counting conventions treat expenditures on con- sdon of the boundary of production in the housesumer durables as a part of (current) consumption hold context. A significnt issue on both concepexpenditure and not as part of the assets of tual and empirical rounds is related to imputahouseholds. But this treatment Is Increasingly tlion. Many of the basic problems are already difficult to justify (Ruggles and Ruggles, 1986), rcognzed in compiling the national accounts especially for long-lived Items. Furthennore, and, just as there isa long-snding debate about some durable items double up as capital equip- the extent of imputation, so there is considerable ment in household production as well as satisfy- controversy about the most appropriate methods in{ household needs (e.g., vehicles and other of valuation that should be applied. There are, transport equipment). There are two problems however, significant strategic differences in the here. One Is the asynumetry in the treatnent of methods of imputing incomes and expenditures essentially the same good as a capital item in from micro data files compared with those at the production and as a current item in the institu- macro level. Certainly the existing custom and tional account. The second is the difficulty in practice in relation to earlier household surveys apportioning oudays between the two uses and has differed markedly from those of the national the respective accounts. The outcome of both of accounts in a number of key respects (Ruggles these issues materially affects our estimates of and Ruggles, 1986). For present purposes some the various aggregates and, in particular, con- of the issues can be highlighted under two main sumers' exwenditure. categories. A.26 Formally, to complete the system of A.28 Thefirstandperhapsmosttractablegroup household accounts, some recogition should be of imputations relates to the consumption of nonmade of the level as well as the change in marketed goods and the orresponding payment household's assets and liabilities. Household of income in kind. As noted earlier, this is not wealth accumulation (which may be negative) formally restricted to farm produce, although in and asset ownership are also central to individu- practice it does account for most of the imputed ais' well-being. There are significant conceptual output in this category, due especially to the item difficulties here too, not least in deciding how to "consumption of own produce." This type of value assets and how to record capital gains and imputation is readily achieved if a valuation of losses (revaluations). In the United States it has the product can be obtained from the srvey rebeen estimated (Ruggles and Ruggles, 1986) that sponse by the individual household. 3 Of course revaluations account for more than two thirds of the same could be true of certain non-marketed the increase in household wealth over a thirty- servkes, but examples in practice are relatively year period. The process of adjushnent will in- few. evitably lead to marked changes In the cross- A.29 Asecondcategoryof imputationsrelateto household asset distribution, but this will not be those non-marketed goods and services for which covered by the existing methodology. no direct valuation can be obtained from a given household survey retum, but for which an impu- Transactio, imputations and atfributions tation is nevertheless required for comparablity with custom and natonal accounting practiee. A A.27 The eore of any set of estimates of house- classic exampleof this would bethe imputed rent hold incomes and expenditures is represented by of owner-occupied dwellings. Although there is the observed transactions in goods and services sonmtmes an attempt to elicit a valuation diarising in the market economy. 2 The market rectly from the respondent, i is often unreliable, valuation is often subject to a range of distortions and hence it is more usual to impute a rental (e.g, taxes and duties) whose incidence could income based on the charactrstics of the dwellmaterially affect Othe comparability of the esti- ing and knowledge of market rents for compamates across households or household groups. rable dwelling characteristics. Nevertheless, market price ndmsurement tends A30 These are dear-cut cases of imputation, to be the orny feasible option in computing esti- but there also exist instances where there is some mates based on marketed output. But beyond conceptual ambiguity about the appropriate atmarketed output there are non-arketed coktpo- trbution of certain incomes and expenditures and nents which result In both actual and implicit about whether further imputations may be necincomes and expenditures. These were briefly essary. An example is provided by the treatment referredto earlier in connedon with our discus- of employers' contributions to social security, 13

218 pensions and Insurance schemes. There are sev- portance at the mnicro level it is suggested that an eral techrdcal issues here. National accounts attempt is made to impute "use valuese in Hne conventions usually Impute an income and an wlththedepreciationofdurables overtnme. (KCatz outlay to households even though, by definition, and Peskin, 1980, have produced esfimates of households do not receive or make hiese contri- this nature for the United States.) butions. Benefits from pemsion and insurance A.32 A final area of concern In estinatiag agfunds are then treated as a change in the form of gregates of the household accounts is with the household assets; otherwise income is counted important distinction between household contwice. However, at the micro level the usual sumption and nwasures of consumers' expendiconvention has been to count the benefits as in- ture. They are not synonymous concepts, and come and to disregard the employers' contnbu- this is best exemplified by the provision of public tions for household account purposes Clearly, to goodsand collective services. Householdsbenefit do otherwise would necessitate seeking informa- collectively from defense and individually from tion about such contrbutions from sources other the provision of health and education services, than the household respondent. It is an area of unresolved odnceptual difficulty. but both items would be excluded from consmri- ers'expenditure. This problem has been brought A.31 The distinction between current and capital Items in the complete household accounting sharply into focus with the interest in aligning the material product system (MPS) for Eastern scheme is crucial to the determination of the in- European countries with the Westen accounting come and expenditure aggregates. A classic de- system (Saunders, 1980). There is no fundamenbate surrounds the treatment of consumer tal problem about Imputation: such goods and durables (Katz, 1983). National accounting con- services fall within the existing boundary of provention has always been to count expenditures duction so that income is generated and recorded on consumer durables as part of current con- accordingly. But there is a problem about valuasumption, completely written off at the time of tion, and whether the present valuation Sat cost" purchase. But this is difficult to justify in view of is really appropriate in these circumstances. the services that consumer durables provide to However, the main problem is one of attribution. the household over a long period. The distint- These services are not "consumedw by the state tion between dwellings, which count as a capital asset, and durables, which do not, is not clearbut by households, so it might be logical to at- tribute the product as an imputed income and cut Indeed, at the micro level it is not even expenditure to households in accordance with possible to argue that such expenditures smooth their use. Conceptually, it would be possible to out across an aggregation of households. Overall, do this for, say, education and health but infeathe balance of the debate on conceptual grounds sible for public administration and defense. In is now towards a new treatment, but there are practice It is difficult to achieve this for most two views as to how this should be done. Some public services although attempts havebeen made argue that the purchase of consumer durables (Meernan, 1978) which night make this a fruitful equivalent to an increase in stdcks, and the actual extension to explore. As in the case of durables, use should be apportioned over their lifetime. the problem becomes more acute, and yet more Others argue that the production boundary difficult to solve, at the micro level. At the macro shoudbeenlarged,by treating dualesas though level the number of options to facilitate approthey generate consumption services. In the latter priate attribution in the national accounts is so case some allowance has to be made for the de- much greater. preciation of the capital stock in common with all other productive assets, plus an interest oppor- Houseold accouns within an intqrated system tuntty cost element, unlike producer durables. The main obstacle to a solution of the problem is A33 The household acounts havebeen viewed obtaining a feasible estimation procedure. Once as a coherent representation of the household again imputations are involved, and there ought economy, stressing the importance of considerto be no fundamental difference in the conven- ing measures of incomes and expenditures as tions at the macro or nmo levels. In each case part of a structure that recognizes the fundamenimputations have to be based on information tal econonmic activities of production, consumpabout fte levels and changes in the stocks of tion and the accumulation of assets and liabiliconsumer durables. In view of its particular im- ties. The overriding message that emerges is that 214

219 whatever is included as incone and expenditure both the conceptual and practical level in the depends crucialy upon what is defined as prod- estimation of household accounts are either trivw uct, in what sense it is consumed and whether or ial or unimportant. On the contrary, it Isimperanot it is induded in the measurement of well- tive that the estimates of the houshold accounts being in some subsequent period. are constructed in such a manner and level of A34 Equally, the household economy is not, in detail as to anticipate the needs of subsequent general, independent of the rest of the economic analysis based on these data. Sindlarly, plans for systm, and hence the household accounts are analysis must define precise concepts of the not to be treated separately from those depicting household variables they require, subject to the activity in the economy as a whole. This is the constraints imposedby data availabilityand data discipline afforded by the macro-level national quality. accounts, in which similar sets of accounts are A37 It is, of course, difficult to anticipate the also constructed for the other institutions of the needs of all studies for which the household aceconomy (companies, govenmuent, overseas sec- counts estimates may be used, and in any case tor) and which rake the interrelations explicit. different authors may wish to adopt different The neatest and most policy-relevant means to definitions. It is therefore important to incorporepresent this is in terns of a social accounting rate flexibility into the calculation of the various matrix (SAM). However, for the household ac- estimates. Such flexibility may be achieved by counts estimates to be incorporated within an constructing a hierarchical data set relating to the economy-wide data framework such as a SAM, it household accounts, with at least three levels to is important that they are made conceptually the hierarchy. At the lowest level (the "variable compatible with the national accounts estimates. level) would be those survey responses of direct Sone of the discrepancies are noted above; but relevance to the household accounts, each exwhile there are notable attempts to harmonize pressed on an annual basis for consistency. The these practices (Ruggles and Ruggles, 1986), the next level (the 'sub-aggregate' level) would remaing difficulties are fornidable. comprise aggreations of a small number of these A.35 In an obvious sense, one advantage af- annualized variables which are very siilar in fordedby thenationalaccountsisthatitprovides nature. At the highest level (the "aggregate" a consistency framework for data from inde- level) these sub-aggregates woud be aggregated pendent and quite disparate sources. At the mi- into a small nunber of totals, each of which are cro eve oniy very limited consistency checks are conceptually distinct. Thus, for example, at the available; in general one has to rely upon the lowest level one might have the variable "Annual accuracy and integrity of the survey respondents. Expenditure on Rice; at the next level, the sub- Inevitable problems result, and these are mani- aggregate "Annual Expenditure on Staple fested, for example, in the under-recording of Grains"; and at the highest level, the aggregate incomes relative to expenditure. There is simply "Annual Expenditure on Food and Drink" This no check on this, and it is a deficiency that cannot three-level hierarchy could be expanded to ineasily be overcome in micro accounting. A dude more than one internediate level if desired. broader framework makes the extent of the dis- A.38 This Appendix concentrates primarily on crepancies explicit and may possibly be used as a the highest and lowest level of this hierarchy. basis for an attempted elimination of these incon- This is not, however, to suggest that the intermesistencies. diate levels are unimportant; on the contrary, the sub-aggregates ought to be computed as part of The estimation of household accounts from the same calculation process. However, it is difthe SDA Integrated Survey ficult to make general recommendations about what the precise definition of the sub-aggregates A.36 Estimates of the components of the house- should be, and certainly much more difficlt than hold accounts should be viewed as intermediate in the case of thfe aggregates. The choice of subrather than final products of the Integrated Sur- aggregates will need to be country-specific (bevey. In other words, they are of interest primar- cause of differences across countries in the parily as inputs to studies of policy issues at the ticular categories of expenditure, sources of inhousehold level, such as those represented by the come, etc.) and made with more explicit refrence SDA Analysis Plans. This is not, however, to to the particular studies for which the data will suggest that the issues and problems inherent at be used. 215

220 A.39 The feasibility of constucting the entire Fme A.1: Aggregates of he household household accounts from the SDA Integrated curreat account Survey is examndned in this section. Nevertheless, for purposes of analysis some parts of the ac- (3) IcaWs Incomefromemployment counts are more relevant than others, and some m ep n parts can be estimated with more reliability than Hwm 11..g nmploymant re others. For most analyses the most important in- Income from t fomation is contained in the household curent Remirettnd account, so that estinution of the current account Oleincme is of greatest importance. The household pro- TO i ncome duction accounts are also important in that they generate incones which enter into the current b) Empendiue account. The capital account is generally less Cosume' expenditure on food important for analytic purpose, and furthnre, Non-food consumption expediture in many caseserious difficulties arise in estimat- Rmans ptraiond hmeprodutd ing many of its components with any degree of Cowpttn of homproduictie non-food accuracy. Similar remarks apply to extensions of Imputed expenditure of wage inme In kind the conventional household accounts, such as time use accounts and records of household assets Totleenture and liabilities. Nevertheless, the feasibility of estimating these components of the accounts from the IS is discussed below. tries, is separated into self-employment income from agricultural activity including the process- The choice of current account aggregates ing of crops) and self-employment income from non-agricultural activity. The mnaiing iome A.40 As explained above, the aggregates of the aggeates covernon-factor incomesand various household current accounts should be chosen in misoellaneous Incomes, the majority of which are such a manmer that their constituent components ill-defined. Remittances received by the houseare conceptually similar, but are distinct from the hold as current transfers from other households conponents of other aggregates. A second guid- are separately identified as an aggregate; the reing principle for the choice of aggregates is that maining incomes, which include non-factor inthey should confonm as closely as possble to come eceived by the houshold as current tsaggregative measures which might be used in fers from governent, inome from insrance analysis. Taking these principles into account, andpensionschemesandvanousotherunwearly twelve aggregates are proposed, as set out in identified sources of income, comprise the sixth Frame A.1; six of these are aggregates of house- income aggregate. hold inwomes, and six are aggregatesof household A,42 On the expenditure side the main concepexpenditures. 4 tual distinction is between monetary transactions A.41 To some extent the choice of the aggre- and imputations. The monetay transactions are gates rlects some of the general conceptual is- subdivided acording to the type of expenditure, sues discussed in paragraphs A.04 through A.35 thus giving food expenditure, non-food expendiabove. On the income side the distinction be- ture (including serices as well as goods) and tween factor and non-factor income is made ex- remittances. The last covers transfers that are plcit. The factor incomes are distinguished ac- simnilar to the analogous income aggregate, excording to the factor remuneated, thus giving cept that in this case the curent tnsfer is being income from employment (returns from labor made from the household in question to other supply), rental income (returns ftom the owner- households. The remaining expenditure aggreship and supply of capital, including land) and gates are all imputations, each of which has an self-employment income (a pint return to labor equal counterpart embedded within the income and capital supplied by the household in which aggregates. Thus there is the expenditure the separate contributions of labor and capital counterpart to wage income paid in kind and the cannot easily be Identified). Self-employment value of domestic consumption of the output of income, being by far the most important overall houseold production enterprises, the latter being source of income in the majority of African coun- subdivided between food and non-food con- 216

221 sumption in parallel with the monetay expendi- in which the respondent was self-employed. Intures. come from employmt rilates omly to the for- A.43 Given these aggregates, totals of incomes mer, that is, to situations in which the Individual and expenditures nay be calculated as the sum was an employee of an hsdtution other than the of the respective aggregates. The difference be- household in question. ' Ihe same section of the tween total income and total expenditure is, of questionnaire may also be used to prvide estcourse, household savings over the year. This mates of self-employment income, which in tum savings estimte also appers as a receipt in the may be used to construct altemative estimates of household capital account and is used to finance incomes from household enterprises. the household's net acquisition of financial and A.47 Focusing initially on wag emowryyment of physical assets. situations, the incme frm employment aggrgate includes not ordy the obvious elemnat of Estimaton of the current account aggregata wage payment in a monetary form, but also any bonuses, tips, gratuities, etc., received in the job(s) AA4 It is necessary to develop the above very in question plus the value of payments received cursory explanations of the aggregates chosen by in kind (e.g, payment in the form of food or the defining them precisely and relating the defini- provision of subsidized housing). As explained tions to the infomation available in the survey. in paragraphs A04 through A35 above, such "in Such an exercise is necessarily one of compro- kind" payment should be imputed as both inmse, trying to condorm the avaiable information come and expenditure because in its absen the as clsely as possible to the idealized ex ante household would need a higher monetary intheoretical concepts. In addition, difficulties arise come in order to be able to buy the quantity of the in clsifying and inteting the responses to goods or services suppied as wage payment. partculr questions and in identifying instances A.48 Respondents are asked to vatze these in of doubl-counting. 'e forme problem arises, kind payments and, although no explicit instrucfor example, with some of the 'miscellaneous tion is given, premably do so at market prices. incomes"inknowingwhethetheresponsesrelate Valuing payments in kind at market prices is to current or capital receipts or to both. As an appropriate either if the individuals in question example of the latter problem, consider remidt- have a right to sell the goods and services to tances paid in kind, which will in general be otes on the market or if the quantity they r- classified as an expenditure. However, there ceive is less than or equal to the quantity the would be double-counting if the acquisition of household would itself choose to consume in an the remitted goods was included in one of the unconstrained situation in which al incomts and other expenditure components. Finally, concep- expenditure are in money terms If the quantity tual difficulties arise in classifying items at the is geater than the amount the housemold would marln, and this has the consequence that even freely choose, and the members do not have the the idealized concepts are not uncontroversial, right to sell the excess, then the situation is one of A.45 The detailed suggested procedues for the consained consumption In which the market calculation of the aggregates are set out in the price overstates the benefit of the payment to the Anne This section concentrates on the defini- household. Unfortnately, there is no means of ion of the aggmates and on discussing the main identifyin this situation and, even if it could be conceptual issues and problems involved in their identified, no mens of correcting for it. calculation from the survey. A.49 For each individual, and for each of the five omcupations which relate to wage employfr4ool AGaAxS ment, income fron employment is therefore cal- (i) Income fiom employment and self-emplment culated as the sum of monetary payments, pay- A.46 All individuals in the responding house- mentsinkindattheirmar1et valuesand bonuses hold aged 7 years and above are asked about received. One issue which remains is the treattheir economic activities In the past twelve mnut of taxes and sodal security contributions months, covering up to five occupations in total. relating to wage employment. In the case of the These responses form thebasis for the estimation two main occupations, respondents are given of nome from employment; however, it is first flexibility by being able toquote thr necessary to distinguish occupations in which payments either pre- or post-tax, so that he rethe respomdent was a wage employee from those sponses will presumablybea mnxtre of pr- and 217

222 post-tax rates. Cearly, all e loynent inwome pations in which the respondent was self-emmust be expessed on a consi# tent basis, eiter ployed, information is collected only on the monpre-tax for all households (wi h the associated etary payments received. Bonuses tips are liy tax payments indluded among te expenditures) to be zero or small in most instances of selfor post-tax for all households. The former ap- employment, and payment in kind is most likely preach is preferred because of its greater flexibil- to take the form of consumption of own output, ity - the pre-tax income can be ctmi:d into on which information is collected elsewhere. The post-tax income if required. questionnaire also does not inquire as to whether A.50 However, estimating tax and social secu- quoted money payments in self-empoyment ocrity contributions paid by those employees who cupations are pre- or post-tax; it fe a ;.med here quote their wages post-tax is far from straightfor- that the former is the case and/or th7 '..e majority ward, and, in particular, there is no information of individuals do not pay tax on self-employment in the survey enabling this conversion to be car- income. ried out. Although a question is asked about A.53 Both income from employment and selftaxes in the IS prototype questmonnaire, this is of employment income nuy be estinuted at the inlittle value for the present purposes; it relates to dividual level (unlike the other aggregates) by taxes paid by the household rather than by indi- summing over all their relevant occupations. viduals and tmy cover other taxes in addition to Household-level esimates are easily calculated those on employment income. To estimate pay- by aggregating over all individuals in the housements at the individual level requires detailed hold. Information on the tax and social security system A.54 Self-enployment income should not be in the country in question. Estimating pre-tax viewed as a complete aggregate In Its own right, income in this manner is very complex as the as it does not include payments in kind (concircumtances and characteristics of the house- sumption of own production). However, as exhold must be taken into account, and in general plained below, It may be used to calculate alterthere will not be a one-to-one correspondence native estimates of the incomes from household between post-tax and pre-tax incomes. These farm and non-farm prduction activities. problems could be circumvented by adding to the questiomaire at the implementation stage a (ii) Housewold agriculturalinconme question asking those respondents who indicate A.55 The obvious approach to the estimation of that their quoted wage is post-tax to estimate the household agricultural income is to build up the tax and employees' social security payments they agncultural production account, the balance of have made. Such a question is unlikely to be which gives the estimnate of this income compoespecially sensitive as it would only be asked of nent. A schematic representation of the agriculthose respondents who have already implicitly tural production account used to derive this estiadnitted paying tax. Finally, it should be noted mate is set out in Frame A.2. While the revenues that employers' social security contributions are and costs components of the account are not of not added to the employment income figures to direct interest from the point of view of current avoid double-counting, because income from so- account aggregates, they are of importance for cial security and unemployment benefits is in- other purposes and ought therefore tobe included eluded elsewhere (see paragraphs A.27 through among the sub-aggregates calculated. A.33). A.56 Consideringthe items on therevenueside, A.5M In the case of the third and other occupa- revenue from the sale of crops is estimated totions, respondents are not asked whether their gether with imputations for the value of crops wages are quoted pre-or post-tax. Given that the paid to factors or kept as inputs (e+, seeds); maority of such occupations is likely to be in the these imputations are also icluded among the informal sector, it is reasonable to assume that all input costs. Proessed crop products refer to wages are pre-tax and/or that the respondents food products subject to further transformation are unlikely to have to pay taxes. after harvesting, e.g., maize flour or shelled A.52 As pointed out above, the information in groundnuts. The raw mnaterial inputs for prothe employnent section of the questionnaire may cessing may be supplied by the household itself also be used to build up estitmtes of self-employ- or from other sources. The component "Oher meat income. The procedure in this case is more Agricultural Revenue" includes mostly revenue straightforward in that, having identified occu- from the sale of livestock products, fishing and 218

223 Fnme A2: The agricultural production account In schematic form Rmnues Revue from sale of cops Revenue from sale of proced crop products Other agdcultural revenue Value of consumption of home-produced food Balance. Revenues -Costs = Househod Agricultural Ince CAst ExpenditureoncropInputs Expenditure on input for crop prossing Expenditure on livestock inputs Expenditure on land Depredation of agricultural equipment hunting, plussome nsceuaneousltems,eg.,sales household mean is unlikely to suffer from seriof palm wine. Finally, consumption of home ous bias. On the other hand, if the variable in production is included among the revenues, val- question is erratic throughout the year, because ued at farm-gate prices (the opportunity cost). of seasonality, for example, the cross-household A.57 The cost side of the production account variance may be overestimated. This problem of comprises both factor payments to non-house- variance overestimnation is potenally serious (in hold members and raw material purchases, theory at least) for studies looking at distributhough without explicitly distinguishing them (if tional issues. (See Box A.1, overleaf.) this distinction was required, it could be achieved A.60 The imputed revenues and costs corre- - up to a point - by judicious choice of the sub- sponding to the use of agriculntral output as aggregates). However, in some instances it is not factor payments or as inputs are based on the clear how a particular cost should be classified. quantities involved as well as the opportunity Payments made in the form of crops are given cost. The other major imputation, consumption imputed values equal to the imputed values on of own agricultural output, is based on grossing the revenue side. Otherwise the costs are mostly up the experience over a short period, making self-explanatory. Expenditure on land includes allowance for the fact that consumption may be both rental payments and the value of crops sup- seasonal and so only taking place in certain p-ied by share-cropping households. One ele- months of the year; the necessary valuations are ment of costs is, however, not captured by the supplied by the respondents. Countries conquestionnaire: expenditure on crop inputs for ducting multiple visit surveys will hopefully obprocessing in the situation where these are pur- tain more accurate estinates (ess recal error) for chased from another household. this component. A.58 Note that only current costs should be in- A.61 The remaining elements of the agricultural cduded in the production; the purchases and sales production account are directly estimated from of assets, including agricultural equipment and the questionnaire, with the exception of deprelivestock, belong in the capital account. Never- ciation. The balance of the account gives an estitheless, depreciation of capital assets should be mate of net household agricultural income. inluded in the current account. It is necessary to However, as observed above, it is possible to use make provision for depreciation; this need is re- the direct estimates of self-employment income flected by expressing household agricultural in- to construct a second estimate of household agricome net rather than gross. cultural income. Clearly, it is necessary to use A.59 The estimation of these elements of the the industry code of the self-employment occuagricultural production account is relatively pationsto select only those relating to agricultural straig&-orward, as descnrbed in the Annex. In activity. Household agricultural self-employment many instances the recall period on which the income is calculated as the sum of irndividual annual estimates are based is short, namely the self-employment Incomne over all individuals in interval between the enumerator's two visits the household; individual self employment in- (typicaly two weeks); the consequences of this come is the sum over all agicultural self-emneed to be noted for purposes of analysis- The ployment occupations recorded for that indirecal error may be quite small and, although vidual. large errors ray occur in grossing up for any one A.62 A serious difficulty in using this esimate individual household (Scott, 1988), the cross- of household agricultural self-employment in- 219

224 Box A1: Depreciaon of productive capital assets Mhe purcdase of capital equipment for use in household be estmated given that two o-ordinates on the depredaproductm activities represents e acquision of a du- tion curveareknown If thevalueofasset j iscuretly V; iable asset with a Ufetime of sev yeas, and so should (in real terms) and was V whe puwchased t years ago, drly be accounted for as an outly in the houseod then its depreciation rate tmay be estmated as capitalaccout However, throughout the pedod in which the hous told owns the asset its value in real terms will V n d ao deteriorate, rrespective of whether or not it is used In Si t production. Tisdeterioration In value, or depredaton. is the result of wear and tear phl possibly some technolovi- e/ cal obsolese. lhus, capital wnsmption is owuring l 4) (geometrcdepredaton) and some allowance should be included In the household cl curent acountl In practice this b acheved by Includg I mra geometric depredaton woud seem to more the value of depreciation among the current costs in the dosely approximatereaity, as onecannotefrom theprcw producuon accunt. Consequently, the balance of h of second-hand capital ises of different vintags househld curment acount becomes net saving that b In this way avalueof Scan beestimatelforeachassetof saving ater provision for depredation or a measure of the type jpesd by ueyed households. However rather additional capital resources over whdch the household has than using a dfferent depredation rate for each individual free disposal while mantining its capli stod& Intact. asetof a Om assettype ftmaksresene toessmte the esvltu on of deptedatnon t btaw d on the facs that an aveage depredation rate over all households for each thevalnof theassetsknownat twopoinee wosesepara- asset type, E say, whic will be used to esmate the value donin t ImeJ ownl Oncethese valum wesaxpressed in ofdd;oforallassetsincategoryj lhissparticlady real tewn it Is then necessary to make an assumption advisable given the difficulty rpondents wi have in about the time proffie of depredation. Two common as- ar ingaccuratetyqisonsonpurchasevaiues. Given sumptins are: i the value of depedation of an _ in category i D O Straight line depredation, that the absolute value of may be eston as ared catal consumption Is the same each year (i tetme)d maat 01) GeometrI depredaton, that depredation occtus to the v same proportionate extent each year. D I - x CUrrent value of asset Whichever assumption is made, the depredation rate may come is that it is not clear what elements of the on estimates of self-enployment income from production account this estimate encapsulates. non-agricultural activity. But a third esimate As it relates to monetary payment only it will not can be constructed on the basis of the response to include consumption of own output, and It is direct questions on profits of the eterpis higly unihkely that it wil be adjusted for depre- A.65 Questions in the survey relate to up to ciation of equipment. The assumption is that al' three enterprises owned and operated by the the remaining elements are accounted for in the household. For each entprise, production acresponses in the employment section of the Inte- counts can be built up along the lines of Frame grated Survey (except the imputed elements A.3. The contents of these ace,unts exacldy parwhich appear on both sides). allel those of the agricultural productionaccounts; A.63 In this manner two estinates of net house- in addition to explicit costs and reenues in cash hold agricultural income may be made. Order of and kind, the value of consumption of own output magnitude comparisons may be made between is included among revenues (the counterpart bethem to check their consistency, though the am- ing included among the expenditure aggregates) biguity in the definition of the second estimate and the value of depreciation of capital assets means that rigorous conr;vfisons camot be made. among costs. The balance of these accounts gives the net profit of the enterprise, and the aggregate (i) Non-farm slf-empoent incom is calculated as the sum of these net profits over A.64 The estimation of net non-farm self-em- all enterprises owned and operated by the houseloyment income is computed in a manner similar hold. to the agricultural equivalenl Thus estimates A.66 Estimation of the revenue components is can be made by onstructing the production ac- based on grossing up the experience of a twocounts of the businesses and alteratively based week period, using specific recall for enterprises 220

225 Frame AM: The production account of a non-fam enteprise Revenue received in money Revenue received in kind Value Of doaectic COnsUmptIn of output Curent expenditure on Inputs Depreciation of capital assets lalnce: Revenues - Cost = Net non-farm sf-employment inome which have been operational in the period since prises not operating in the period between the the enumerator's last visit and "normative' esti- enumerator's visits, and the more flexible apmates based on typical experience otherwise. proach to the collection of expenditure data means Current input expenditure is estimated in a simi- that the resulting figures are likely to be more lar muaner, except that in this case respondents reliable. On balance, however, it seems that, alare given greater flexibility in the reference pe- though the accuracy of the estimate may be inriod of the response according to the frequency creased (that is, its mean square error reduced), it of purchase. In both cases it is necessary to take is not self-evidenthat the bias will fall sigrdfiinto account the fact that the enterprise may only cantly. Consequently, it is important to derive have been operational for a fraction of the year. alternative estimates wherever possible. A.67 The balance of the production account A.69 A second estimate of household non-farm provides an esimate of net non-farm self-em- self-employment income may be constructed in ployment income. While conceptually correct, in an exactdy analogous manner to the agrcultudal practice, estimates based on this meffiod were estimate, based on all self-employment occupafound to be very unsatisfactory in the Living tions except those in agriculture. Finally, a tdrd Standard Measurement Study (LSMS) question- estimate can be constructed based on the diet naires; in each of the countries concerned the vast responses on profits, comprising the sum of profits majority of enterprises appeared to be making used for own and other household purposes, substantial negative net profits. As such the re- transferstootherhouseholdsandsaving. Inboth sults were not credible, and it was clear that either cases, while it is not clear exactly what elements significant overstatement of costs and/or signifi- of the business account this figure is meant to cant understatemnent of revenues were occurring. represent, it may be assumed that both of these There are a number of reasons why this might figures cover only cash revenue and hiput exoccur. double-counting of expenditures made penditure;consequentlyrevenueinkindplusthe jointly for more than one enterprise or of con- value of domestic consumption of output is sumption expenditures and production input added, and the value of depreciation subtracted. expenditures; capital purchases being inadvert- A.70 As with household agricultural income ently included among current expenditures (for above, it is probably appropriate to make ordy example, under the "Other Expenses" category); order of magnitude comparisons between the alrevenues being quoted post-tax while the taxes ternative estimates. The definitions are too vague paid are included among input expenditures; and to enable precise and detailed comparisons to be traders quoting their margins rather than gross made. revenues but sti reporting the expenditure on the items purchased for resale. Most obviously, (wu) Income from rent revenues maybe deliberately understated because A.71 In common with most of the other income of fear of taxation. aggregates, rental income comprises both mon- A.68 To some extent these problems may be etary and imputed elments. Rent accrues to endemic, but it is likely that they could be re- households through their ownrip and supply duced by better design and admirdstration of the of assets of various descriptions. Considering survey. It is difficult to tell ex ante whether the actual rents (in money or in kind), the question- SDA Integrated Survey is sigrdficantly less likely naire permits estimation of income from the to suffer from bias than the ISMS questionnaire. leasing of the following assets: land (by means of Cerinlyamorecomprehensiveattemptismade either rental or sharecropping arrangements), to collect information on the revenues of enter- draught animals, agricultural equipment, assets 221

226 as the household), to which the questiomaire Box A2: Estimating the Imputed rent of does not refer. While It is conceivable that the ownaeocpled dwellin respondent would include such income among miscellaneous 'Other Incomes catega oies, there hbitantsof ownar-occupied housing as to owners of a isano itsmof knowingk whether thisc hsoccue,d productve aet fiom which they derive consumption and it Seem5 unlty in pracfice. Hence, in value. As such, an imputed mre restingthepoten- modifying the questionnaire for implementation, tia mrket rentl value of the dwelling Is added to both there is still a need to include a specific question the income and expenditure sides of the curent ao on rental income from this source (in section 7 or CoMt. 11, forexample). Unlike some of the LSMS questionnaires, the IS A.72 To these actual rental InoDmes (some of prototype questionnaire does not include explicit ques- whch my be ceived in kind) should be added 1ions on the powtial rental value of owner-occupied thimpute rentlvalue of o eoccupied dwellings. Hence, Lmputed rent must be estimated in- the imputed rental value of owner-occupied direy based on a hedonic relationship estimated for dwellings. Estimation of this imputed rent should rented housing, rdating the annual rental value to the be based on the characteristics and amenities of charerstcs and ameadties of the dwelling. Such an the dwelling. Given this estimate, the aggregate equation may be used to impute rental values to owner- rental income figure is simply the sum of the occupied housing that is, to use their dhuactedstcs to actual and imputed components. (See Box A2_) predict from the equation what additional Income and expenditure would be necessary In order that the household could rent a dwelling of comparable d.7a3- (v) Remittances r eaei ed tetdstic. A.7 Remittances received relates to aent The appropriate explanatory variables are mostly transfers from other households, which may be contained in the Housing Section of the Integrated Sur- in the form of money or in kind (in particular vey, such as dwelling size, quality of construction, ac- food). The aggregate is the sum of such transfers ce to amenies; characteristics of the household may received in the past twelve months. The only also berelevant, forexample if adiffoent rent is charged issue which anses in its calculation is the need to to eigners. Although some of these variables are continuos, the majority are discrete. It Is therefore neces- distinguish current and capital remitances. In sary in the latter instances to construct dummy vari- the case of capital remittances, payment of the ables (e.g, a variable taking the value 1 for a 'good transfer implies an economic obligation for requalityconsuctionofthewausand0for'lowquality payment; that is, from the point of view of the cotrucion The equation may be estmated by ordi- recipient household, a financ liability is estabnary least squares methods, although in this case there lished. Capital and current remittances can be is a possibility of sample seletion bias (Maddo n9s distinguished using the questionnaire; the forcli.9). Tisbisbecause the estimatedequation represets the prefens of renting households, but is used to mer should of course be included in the houseimputerenta valuesfor thedwellings of owner-occupi- hold capital account. es Housebolds who are owner-occupirs may have dirent preferenees from renting househods (e.g., in (vi) Other income Afria the fanner are more likely to be rural and the A.74 Thisis an amalgamation of various, mainly latter to be urban). In practice this bias may not be non-factor, incomes. Many of the transfers in Important and In any case may be small relative to the question are received from governmt, eg., unother errs in this procedure. A possible difficulty with this prooedure for imput- employment benefits and educational scholar- In$ nnts to owner-occpied dwellings may arise in ships. A number of the items of miscellaneous ruralraeas ff there are insufficent rented dwellns to receipts are hard to classify at the margin, ie., enable a proper spedfication of the bedonic equation. If whether tiey should be included in the current this is so the consequences tay be quite serious, as the or capital accounts. For example, should income majority of owner-ocupied dwellngs are in rural ar- from dowy or inheritance be clasfied as an ea. To avoid this possibility there is a strmg cf infreqluent, nn-factor income or as a capitl reincuding direct estimates of potential rental value in the quesdonnaire, the estimates being provided either ceipt? (The former assmption has been made by the respondns or by the enumertors here.) The contents of the "income from other sources" questions in section IIC are partictuarly difficult to classify. of non-farm enterprise. One obvious omission A.75 Ingeneral,thereisadangerofunderstateis the inome from renting dwellings (including ment of miscellaneous incomes, due to the fact income fmm tenants living In the sane dwelling that ony a few categories are included in section 222

227 Ftame AA: Schematic representation of the household capital account Saving (balance of arent acwount) Balanem Zero NOW Th capi a Int aprestd hekm indudes he capit iane act. Net acquisiton of physical assets land Livestock Agricultual equipment Businesassets Duablegoods Inventode Net acqwsition of findal assets Repayment of loans Capitalremdttac" Paid AC and some of these are not very specific. Con- recall period. Finaly, in countries underaking sequently,respondentsmayforgettoreportsome multiple-visit surveys, expenditue between the iscelaneousincomesintheresidualcategories. first and last visit could probably be estimated There is a case for including additional specific more accurately than the single recall, the resultquestions on items such as the following, interest ing estimate being grossed up in the same manon bank accounts, dividends incone from gamblin. Wbile ordy a very small number of housener. (u) Consumption of home production: food holds will have these income sources, they may A.78 Consumption of home-produced food is be significant for those who do. presented as a separate aggregate, calculated in exactly the same manner as on the income side A.76 Exmcm=AG=Ans. Inforationon the (see above). main items of household expenditure iscollected (m) Consumption of home produdion non-food by the survey at a highly disagegated level and A.79 Similarly,consunmtionofhome-produced With flexibility in the reference period. Both of non-food items also appears as an expenditure, these factors, plus the fact that there is less in- again calculated as described above. centive to understate because of fear of taxation, (iv) Non-food cownputi expenditure should imply that, in general, the expenditure A.80 This aggregate includes a wide range of r.stimates will be more accurate and less subject goodsand services. Moredisaggrgted estimates tdbiasthanthe inomeesfimates (though itshould wil be requied for many purposes, and this be remembered that some elements are common should be reflected in the choice of the sub-agto both sides of the account). gregates. The majority of the non-food items are (i) Cwnes' e2penditure on fod included in the household expenditures module A.?7 Food expenditure (that is, market pur- (section 9A). The estiation of expenditure on chases of food) is estimated by summing annual anyitemisconducted ina similar manner tofood expenditures at the individual commnodity level. items (see above); as with food expenditures there The latter is estimated from information based is a variable reference period according to the on fairly short recall periods: the interval between frequency of purhases, and for the most frethe two visits for frequent purchase and the pre- quently purchased items the number of months ceding three months for less frequent purchases. in which, not consumed is also recorded. Annual estimates are derived by grossing up in A.81 This section, however, does not cover all the usual way, taking into account the fraction of the expenditures of the household on non-food the year in which the item is not purchased. Thus items; specifically, the following Items need to be the responses are based on recall of actual experi- estimated from elsewhere in the questionnaire: ence (rather than "typical figures), but the po- expenditure on housing (rent, imputed rent of tential problem of seasonality which gives rise to owner-occupied dwellings and repairs or mainoverestimation of cross-household variance is tenance costs), expenditure on household utilirendered less serious due to tie flexibilty in the ties (water, electridty, etc.), expenditure on edu- - ~~~223

228 Box A.3: Estimating consumption flows from durable goods sion of the latter may mean that some doublecounting is introduced, but thee is no means of checking for this. (See Box AW3.) us may pprod eitaen t Oms o A.82 The rensaining Items are slightly less the onsumpption flows, or 'use values," a houho straig dar As explained in paragraphs A.04 receives by virue of Its owneshp of a durable om through A35 above, acquisition of durablegoods suenr good (Katz, I953). One apprii b to pn has been treated as the purdhase of a capital asanalogously toth esmation of imputed rents of ownr- set and as such is included as an outlay in the occped dwellins, tat is, to bas the estimate on the capital awcount. The durable goods are, howmarket rental value of the good. However, in price ever, perceived as generating a flow of consunpthe ne ry Informatin is not usually available to tion sevices to the household which owns them, make ths approadc operational, and In any case con- estmated based on depreciation and the interest sumer durable goods, even of a particular type, are generaly very beerogenous in natur An alteratlive, opportunity cost. Finally, with the m morefeaslbleapproachisbasedantheopportunitywst ous items of expenditure in section lld of the of owning the duable good, In gera ths wil be questionnaire, a problem arises analogous to that comprised of two elemeits ocurring with miellaneous incomes, that is, (l) the value of depredation over the year in question; the distinction between curent and capital purad chases Presumably conthbutons to self-hp GO an imputed net retun, that Is the maximum rate f prchss will be apitbl "ouaysn but it isless clear rethn which could have been obtained on some aternativeivebent how to classify 'gifts and presetsw or 'wed- The depredaton rate, 8 is calclated in the same dings,dowly, fnierals etc." One of the categories, mannerasfrproduoerdwrales. hecalculatlonof the 'expenditure on trwportation, is covered elseinterest opportunity cost, r, is nore probematic. In where in the questonnaire and is in any case Africa many mrl houseds nay not effectively have ambiguous; respondents will not know whether aa to interest-bearing savings acwounts and othe it is intended to relate to consumption or producfinanca assets, and so the apportudty cost for such a.. ( houwhdmaybezem FormnanyhouseholdsInutban tion expenditure (or both). areas there mnay be a genuine opportunity st, but it (a) Reittanes paid out wflbedifficulttoidentifysuchhousehodsaswellasto A.83 This aggregate includes payments in deermine the appropriate interestrate. If this could be money terms and in kind; as with the cowedone, and assuming gometc depedaion, the con- sponding income aggregate, it is important that sumpton flow, cy from a durable good of current value only current transfers are included. In the case of v, may be estimated as remittances paid in kind there is the possility of * vi, (i+ r)/n - 5some double-counting if the expenditures (actual Vi. (c+"al or imputed) involved in acquwi the remitted with r being set to zero in approprate stnces. As goods are quoted elsewhere in the questionnaire. disued above, it may be necmssary in practioe to (vi) Imlputed expndit u of wage icome i kind assume r is zer for sa households. A.84 This is an aggregation of the various forms of payment in kind received as wage incom in each of the occupations undertaken by each cation, the imputed flow of consumption from member of the household. As with remittances, durable goods and various miscellaneous expen- it is assumed that these imputed expenditures ditures. Imputed rent of owner-occupiedwell- are not covered by the other expenditure aggreingsis included in expenditure on housing as the gates. counterpart to the identical item on the income A.85 This completes the estimation of the side (see above). Expenditure on utflities covers household current accounts, summarized in the water, electricity and garbage disposal; annual twelveaggregatesreferdtoabove. Clearlythere estimates are derived by grssing upon the basis is no single unambiguously correct definition of of the most recent bill. In the case of the water bill these aggregates, and diffeent users may wish to it is necessary also to take into account private calculate slightly different aggregates (in parpurchases of water from other households and to ticular, perhaps differing in the treatment of subtact the value of income from private sales to marginal items); this may be achieved by making other households. Expenditure on education slight modifications to the procedures Listed in comprises both direct costs (such as fees) and the Annex Clearly, the agegates also need to indirect costs (such as tansport toschool); inclu- be supplemented with a large number of more 224

229 deal and specific sub-aggregates. outstandingloans. A.89 Unfortunately, not all of these capital Capitl accoun, assets nd labies and time use trsactions can be estimated from fte information provided by the IS. Purchases and sales of A.86 In addition to the informadon on incomes the folowing assets are recorded: land, lvestock, and expenditures, as discussed above, the Inte- agriculturl equipment and the busin assets of grated SurVey also colects infomation on the non-fann enterpries. Estimates of depredation capital tranactions of households, on household are available for the last two categories of assets, asseb and liabilities and on time use of house- and ft is not unreasonable to set the depredation hold members Thus, the issue arises of whether of the first two assets to zero. Thus net acquisiit is possible and worthwhile to construct a com- tion of assets in these categories may be estiplete set of household accounts comprising, in mated. However, neither net acquisition of duaddition to the cunent account, a household rable goods nor stockbuilding (except for a limcapital (including capital finance) account, a ited number of agricultural staples) can be estirecord of begingand end-of-perodassetsand mated from the information provided. Turning liabilities and an account on time use for indi- to financial assets, net capital remittances paid vidual household membe. However, should out tootherhouseholdsmaybestraightforwardly suchanambitiousobjectiveprovetobeinfeasible, estimated once cunrent and capital inter-houseitwfll still generallybedesirabletoestimate some hold transfers have been distinguished. Howof the components for policy purposes (eg., to ever, a much greater difficulty arlses in the estiidentify households whih are net purchasers and mation of net loan repayments, because responnet sels of land). dents are not asked to separately identify repay- A.87 Considering initially the capital account, ment of principal (which should be included in which is represented in stylized form in Frame this item) and interest payments (which ought to A4, household savings is estimated as the bal- be indluded in current account tansfers). anceof the household current account. Although A.90 The fact that complete capital accounts there are direct questions on household savings cannot be constructed for households implies that over the year, these rlate only to savings in for- it will not in general be possible to build up a mal finandal institutions. Having to estimate completerecordofhouseholds'holdingsof assets savings as the balance of the current account is and liabilities at the beginning and end of the problematic, in the sense that any errors which period. The picure is, however, more complete affect one side of the account but not the other on assets than on liabilities The value of the wil cumulate in the saving estimate. Further- followingassetsowned by the household maybe more, the capital account is unikely to balance esfimated:land,lives,agculturaleqipment, Hence, the savings estimate which resltsislikely stocks of key agricultural staples, assets of nonto be highly inaccurate, and, to the extent that farm enterprises, savings held in formal financial incomes are likely to be underestimated signifi- institutions (for each of which the value at the cantly more than expenditures, it is likely to be beginning and end of the year may be estimated) downward biased. and consumer durable goods (for which only the A.88 Inthecapitalaccount,savings(whichmay end-of-period value may be estimated). On the be negative) are abocated either to the net acqui- other hand, there are a number of assets which sition of physical assets (net new investment in cannot be valued, such as stocks of most agriculfixed capital or inventories) or of financial assets, tural products, the value of owner-occupied either or both of which may be negative. Because housing and of informal sas Unfortunately, of the treatment adopted above, purchases and liabilities cannot be estimated at all; housholds sales of durable goods are regarded as positive with outstanding loans can be identified, but the and negative net investment in fixed assets re- value outstanding is not known Thus, in sumspectively. And the fact that profits of household mary, as with the capital account, only a partial selff-employment production activities are mews- picture of households' ownersip of assets and ured net means that net acquisition of physical liabilities can be assembled; however, this does assets in the capital account should be measured not mean that the estimates which can be made after depreiation. Pmally, net acquisition of fi- are worthless. In fact, those assets which can be nancid assets includes net capital remittances valued are some of the more important from the paid out by the household and net repayment of perspective of SDA analysis (and, livestock) in 225

230 which, for example, the extent of asset ownership been identified earlier in the questionnaire as might be used in conjunction with other vani- non-applicable and so skdpped. In other words, ables in dassifying households into groups along the response Is not missing but zero. These two socio-econonic and other lines. types of non-response must be distinguished, be- A.91 Similarly, in the case of time use accounts, cause their interpretation and treatment is quite only a partial picture can be built up, this being different. Fortunately this may be achieved by based on the experience of individuals In the carefully following through the "skip patterm t in seven days prior to the interview. As before, the questionnaire in order to dentify individuals however, the information which is available re- of whom the question was not asked; in such mains of interest for several purposes, such &.i cases the response to the question mustbeexpllcmeasuring the opportunity cost of older children itly set equal to zero. Note, however, that this attending school. For each individual the time procedure does not necessarily identify all zero devoted over the past seven days to the follow- values. For some households the skip pattern ing activities may be estimated: the primaiy oc- will ensure that they are not asked any questions cupation of the week (for those who were em- in a particular module; for example, households ployed), the time devoted to collecting wood, who indicate that they do not own any non-farm collecting water and housework. The time de- enterprises are not asked any further questions voted to activities in a typical week can also be related to this topic. Such households will not estimated: time devoted to the more important have a record in the questionnaire section on economic occupations, travelling to work (main non-farm enterprises: absence of a record indijob only), travelling to school and the number of cates zero values for questions of direct relevance hours of school missed. In addition, the time to the estimation of household accounts aggretaken by some one-off activities such as attend- gates. ing health consultations is also known. In gen- A.94 Having identified the zero values using eral, therefore, it is clear that an individual's time the conditional "sldp" pattern, the remaining inuse can only be partially accounted for by using stancesof non-responsemustbegenuinenmissing the Integrated Survey. values. The extent of the problem caused by missing values, and hence the importance of re- Practal issues: Zeo values, missing values ana estimating then, is dependent on the number of outrs missing values (i.e., whether the number of nonmissin& non-zero values is sufficient for the en- A.-92 The discussion above and in Annex 1 sets visaged analysis) and on wiether the households out the theoretical procedures for calculating ag- affected are considered to be a random or sysgregates in the household accounts. However, it tematic sample (the latter might occur if higher needs to be supplemented by a brief discussion income households are more liely than lower of the inevitable and complicated problems which income households to refuse to answer sensitive will arise in practice. Two of these practical is- questions). If re-estimation is considered necessues are endemic, and consequently their treat- sary, the appropriate procedure depends on the ment needs to be made a central part of the calcu- statistical process which is believed to be generlation procedure. These are (i) the problem that ating the variable in question If, for example, it instances of missing values will occur, in which is believed that the variable is nomwally, inderespondents are either unable or unwilling to pendently and identically distributed (perhaps provide an answer to a question; and (ii) the fact within an exogenously determined group of that the responses provided may contain outliers households), then the optinal estimate is the mean which need to be detected and then replaced, of the distribution in question. If, however, it is that is, values which deviate so much from the believed to have been genrated by a general other values as to raise the suspicion that they are linear process relating it to one or more auiliary erroneous. Possible methods for identifying and variables, then the optimal prediction may be dealing with these problems are smmarzed madegiven the estimated equationfor theprocess below. and the values of the auxiliary variables for the A.93 The difficulty in identifying missing val- household in question (assuming the latter are ues is that they are only a small subset of in- not missing). stances of non-response to a question. Apparent A.95 The second problem, the identification of non-response will also arise if the question has outliers, Is controversal but Important; if an out- 226

231 Hier is not identifled as such, and consequently from the SDA Integrated Survey. Neverthdess, not re-esimated, the impact on analysis based on it is recognized that analysts may wish to adopt the variable could be serious. However, extrme sligly different procedures and, in particular, values may arise for genuine reasons as well as different definitions for the variables to be calcubecause of errors, and it is important that these lated. Consequently, a key element of this protwo cases are distinguished as far as possible. posal has been to highlight the flexbility tplicit The procedures for identification of outliers de- in constructing a hierarchical data set comprising pend again on the statistical process believed to variables at three or more levels of aggregation be generating the variable. If the variable is be- but subsumed within a conumon household aclieved to be nonnally, independently and identi- counting structure. catty distributed, then a simple criterion might be A.99 Analysts wishing to use variables derived values lying more than a certain number (eg., from the household accounts either as explanafive) of standard deviations from the mean value tory or dependent variables must define very (both mean and standard deviation being calcu- precisely the concepts they require, bearing in lated orny over non-zero values). The optimnal re- nind the information that is available in the surestimate for values identified as outliers is the vey. A second important issue is the need for mean value (after the elimination of outliers). If analysts to assess the reliability of the estimates thevariableisbelievedtobegeneratedbyalinear they wish to use. While there are no objective relationship with one or more auxiliary variables, means of assessing error available, some printhen standard methods for identifying outliers ciples can at least be set out. Unfortunately, few from regression residuals may be employed, the instances of alternative estimates exist, although outliers being re-estimated using the predicted comparisons between total income and total exvalue of the equation for the household in ques- penditure can be used to provide an order of tio. 6 magnitude assement. The danger of overesti- A.96 The identification and re-estimation of mation of the cross-household variance of an anoutiers should be conducted at each level, start- nual variable estimated by grossing up an estig Initialy at theannualized variable level (which mate based on recall over short periods has been should renove most of the outliers), then at the noted above. Items based on typical values can sub-aggregate(s) and aggregates level, adopting be expected to give less accurate estimates of the the same criterion in each case. Unfortunately, crosshousehold mean than items based on recall however, this does have the effect that in one or relating to a specific period. Also, items derived two instnces thc value of an aggregate may be as a residual, such as saving, can be expected to inconsistent with the values of the sub-aggre- be highly inaccurate. gates and annualized variables. A.100 As explained above, it is not possible to construct a complete and comprehensive set of Conclusions household accounts estimates based on the Integrated Survey. In particular, several items in the A.97 It should be clear by now that the estima- capital account cannot be estimated, orny very tion of superficially straightforward conceptsuch incomplete time use accounts can be constructed as "incomew and "expenditure" is by no means a - and the stock of outstanding household liabilities trivial exercise. Many significant and not com- cannot be estimated at all. This may not be too pletely resolved difficulties arise in defining pre- serious, however, the problems may be endemic cise concepts of income, expenditure and other in that respondents may not be able to provide household accounts variables. At a more practi- the requisite information, and, in any case, these ca level, the information provided by household items which cannot be estimated are generally surveys must be appropriately transformed and not crucial for SDA purposes. The most imporaggregated to build up as close an approxima- tant items are generally those contained in the tion as possble to the desired coneepts. At an household current account, the components of even more practical level, problems caused by which can be estimated more or less from the missing observations and outliers must be tack- questionnaire. However, there are some omisled. sions which could be remedied in a straightfor- A.98 Thischapterhasdiscussedtheseproblems ward way by adding a small number of quesand set out a suggested procedure for conducting tions to the draft questionnaire when it is being this exercise of esdmating household accounts modified for implementation. 227

232 Annex: Det ed methodology fo current aoxnt agp ThediusslonnthettosAppefds xhasstrssed Theconstuentmporntsofeach aet areset that ingenal thue wil be no sinle, unambiguously- out along with an udikttlon of the geric type of correc defnition of measures incomes caculation used in ta variableinto the and expenditures. Ambiguities wil adse both at the mesedqulredfor ntusonh conceptualevel and, much more sinfindy, in de- is a limited number o such genec cation p terminutheconceptswhichrespondentshaveinmind duies; thus teae bythelletea, b... h, when ing monetary reponses to questions. each referig to a categolyof calculation as Nevert*heles a suggested definitonal and nefthodo- defnateendofthlpd Absenceofaletter logical framework-for the calculation of household indicates that the questionnaie maybe used currett accuunt aggregates is set otm below. dewyeywitgoutanyfwthwy. of Eousehold tes (1) lnom fron employment (and self-employentlnewae) A. Incone from employment (a) Main occupation Paymet In mone Value of bonuses received (if addiona to money paymeno Value of paynent in forn of food, crops, animals Value of payment in form of subsidized housing Value of payment in form of subsidized trasport Value of paynent in other forms Estimated taxes and social security contributions paid (if wage quoted post-tax) (a) Income from main occupation (b) Secondary occupation Paynent In money, including bonuses a. Value of payment In fonn of food, crops, animals a Value of payment in other fonms a Estimated taxes and social secunity b conftrbtions paid (If wage quoted post-tax (b) Income from secondary occupation (c) Third occupation Payment in money, including bonuses Value of payment In knd (c) hnome from third occupation (d) Fourth occpation Payment In money, including bonuses Valueof paymentinldind (d) Income frm fourth occupation (e) Fifth occupation Payment in money, including bones Value of payment in kind (e) Income from fifth occupation a' a a a a a a a a a a a *T bbitfs to t demo eoswetyps olcdwaala thata ueps Tabb). doa *tepmdedved e km do IS -x ( the esid of %e 228

233 To calculate (}A) income frm employment, the income from occupations in which the individual in questin was self employed rather than a wage employee must be set to zero. Having done so Income f*om Employment = (a) + (b) + fc) + (d) + (e) To calculate employnent income at a household level thds should be aggregated over al individuals in the household. a Self employmt Income (a) Main occupation Payment in money Value of bonuses received (if additional to money payment) Estimated taxes and social security contributions paid (if wage quoted post-tax) (a) Cash income from main occupation (b) Secondary occupation Payment in money, including bonuses Esfimated taxes and social security contributions paid (if wage quoted post-tax) (b) Cash income fr*n secondary occupation (c) Third occupationr Payment in money, including bonuses (c) Cash income from third occupationa (d) Fourth occupation Payment in money, including bonuses (d) Cash income from fourth occupation (e) Fifth occupation Payment in money, including bonuses (e) Cash income from fifth occupation a a b a b a a a To calculate self-employment income, the cash income from occupations in which the individual in question worked as a wage employee as opposed to being self-employed must be set to zero. Having done so Self-employment income = (a) + (b) + (c) + (d) + (e) To calculate self-employment incomne at a household level this should be aggregated over au individuals in the household. 229

234 (2) Household agdcultural Income (a) Revenue from sale of crops (i) Staple grains, field crops and cash crp Sales revenue through main outlet Sales revenue through other sources Value of crops given to landlord Value of crops used in processing Value of crops kept for seed Value of crops paid to labor Revenue from sale of crop, aggregated over all crops in category (a) (I) Revenue from sale of staple grains, field crops and cash crops (ti) Roots, fruits, vebles and other ows harestd piece meal Estimated annual sales revenue Esdmated value of crops paid to landlord Estimated value of crops used for processing Estimated value of crops paid to labor Revenue from sale of crop, aggregated over all crops in category (a) (.i) Revenue from sale of roots, fruit, vegetables and other crps harvested piecemeal (b) Revenue from sale of processed or transformed crop products Estimated value of sales of processed crops, aggregated over all categories (b) Revenue from sale of prooessed or transformed crop products (c) Other agricultural revenue Value of sales of product, aggregated over all commodities (c) Other agricultural revenue (d) Consumption of own agricultural produce (i) Grains and flours; roots and tubers; pldses, nuts and seeds; fruit Estimated value of annual consumption, aggregated over appropriate comnmodities (d) (i) Consumption of own agricultural produce I (U) Vegetables; meat and fish; livestock products; drinks Estimated value of annual consumption, aggregated over appropriate commodities (d) (ii) Consumption of own agricultural produce II c c c c c c c d e e 230

235 (e) Expeditre on crop inputs Experditre in cash and kind on crop hiputs, aggregated over al int Value of crops used as seeds l: staple grins etc., aggregated over appropriate commodities Value of crops paid to labor 1: staple grais etc., aggregated over appropriate commodities Value of crops paid to labor I: roots etc., aggregated over all crops in categoly (e) Expenditure on crop inputs () Expevditu on inputs for transforming/prcessing crops Value of crops used in processing!: staple grains etc., c aggregated over all crops in category Value of crops used in procwing 11: roots etc, c aggegated over all crops in category lbo costs, aggregated over all products d Other costs, aggregted over all products d (0 Expenditure on inputs for transfonning/procsng crops ( Expenditure on livestock inputs Expenditure on fivestock inputs, aggregated over au input tpes (g) Expenditure on livestock inputs (h) Expenditure on land Payment for renting plots, aggregated over all plots Value of crops paid to landlord 1: staple grains etc., aggregated over all crops in category Value of crops paid to landlord IT: aggregated over al crops in category (h) Expenditure on land (i) Depreciation of agricultural capital asses Esimated annual depreciation of assets, aggregted over all categories of asset.i) Depreiation of agricultural capital assets (p Agricultunl self-employment income in cash Self-mployment income calculated over agricultural activities ondy, summed over all individuals In household (p Agricultural self-employment income in cash Household agriclltural income (estimute 1) =(a)+(b)+(c)+(d)-(e)-(f)-(g)- h)-(i) Household agricutural income (estimate ) (p + (d)- (i) qpes c c c c c f 231

236 () Net non-fam self-employment income (a) Revenue received in cash (i) Enterprise operatig sice last visit Revenue received in cash, aggregated over all enterprses (a) (i) Revenue received in cash I (U) Enterprise not operating since last visit Estimated reveue received in cash, aggregated over all enterprises (a) (ii) Revenue received in cash 11 (b) Revenue received in kind -) Enterprise operating sie last visit Revenue eceived In goods and services, 8 agpepted over all enterprses (b) (i) Revenue received in 1dnd I () Entferprise not operating sinc last visit Estimated revenue received in goods and services, aggregated over all enterprises (b) (ii) Revenue received in kind I (c) Value of domestic consumption of output (I) Enterprise operating since last visit Value of domestic consumption, aggregated over all enterprises (c) (i) Value of domestic consumption of output I (ii) Enterise not operating since last visit Estimated value of domestic consnption, aggreted over all enterprises ic) (ii) Value of domestic consumption of output I (d) Current expenditure on inputs (i) Enterprise operating since last visit Annual expenditure on item, aggregated over all items and al expenditures (d) (i) Current expenditure on inputs I (ii) Enterptise not operating since Last visit Estimated annual expenditure on item, aggregated over anl items a all enterprises (d) (ii) Current expenditure on inputs D (e) Dprepdation of capital assets of non-farm enterprises Estmted value of depredation of asset, aggregated over all categories and all enterprises (e) Depreciation of capital assets of non-farm enterprises g g g g g g g f 232

237 (f) Gross profit of non-fann enterprises Annual value of profit used for own purposes, aggregated over all enterprises Annual value of profit used for household purposes, agreated over all enterpries Annual value of additional profit, aggregated over all enterprises (f) Gross profit of non-farm enterprises (g) Nfarm self-employment income in cash Selfemployment income calculated over non-agricultura activities only, summed over all individuals in the household (g) Non-rm self-employment income in cash g g g Net non-farm self-employnent income (estimate) = (a) + (b) + (c)- (d) -(e) Net non-farm self-employment income (estimate 11) = ( + ) + (C)- (e) Net non-fam self-employment income (estimate 11l) = (g) + (b) + (c)- (e) (4) Actual and Imputed rent (a) Actual rent nme from renting land Incme from giving land out for shaae-aop ncome from renting draught animals Income from renting agricultural equipment, aggregated over different items hincme from renting assets of non-fann enteprises aggregated over all enterprises (a) Actual rent (b) Imputed rent of owner-occupiedwellings hnputed rent of owneroccupied dwelling (b) Imputed rent of owner-occupie dwellings h Actual and imputed rent = (a) + (b) (5) Remittances received Curent remittances received in cash, aggregated over all donors Current remittances received as food, aggregated over all donors Other current remittances received in kind, aggegated over all donors (5) Remittances received 233

238 (6) Oher Income Value of educational scholarship, aggregated over all Individuals Iucome from social security/unemployment benefit hcme from state pension Other incme fon goverment sdurces Income from private pemion/insurance Income from dowuy/inherltance Oteinconm (6) Other income (7) Consumes expendittre on food Annual expenditure on commodity, aggregated over all conunodities (7) Consumes expenditure on food g (8) Consumption of home produced food This is the counterpart of the coresponding income item (see 2(d) above) and is calculated in the same way. (9) Consumption of home-produced non-food items As with agregate (8), this is the countepart of the oresonding income item (3(c) above) and is calculated in the same man. (10)Non-food consumption expenditure (a) Expenditure on education School and registratfon fees Contributions to parents' associations Uniforms and sports clothes Books and sdhool supplies Transportation to school Food, board and lodging at school Other expenses (dubs, extra classes) Other in kind expenses Expenditure on individualrs education, aggregated over au individuals (a) expenditure on education 23

239 (b) Expenditure on housing Rent paid in cash Rent paid In kind Imputed rent of owner-occupiedwellings (b) Expenditure on housing (c) Expenditure on household utilities Annual value of water bills Private purchases less sales of water Annual expenditure on electricity Annual expenditure on garbage disposal (c) Expenditure on household utilities (d) Other non-food household expenditures Expenditure on taxes Expenditure on weddings, dowry, funrals, other cermonies Expenditure on gifts and presents Other miscellaneous expenditures Annual expenditure on household item, aggregated over all categories (d) Other non-food household expenditures (e) Use value of durable goods stmated use value of category of dumables, aggregated over all categories (e) Use value of durable goods d d d d d d g j Non-food consumption expenditure = (a) + (b) + (c) + (d) + (e) (1V)Remittmces, paid out Cumrent remittances paid out in cash, aggregated over all recipients Curent remittances paid out as food, aggreated over all recipients Oer current remittances paid out in kind, aggregated over all recipients (11)Remittances paid out 235

240 (12)Expaendtue coresponding to employment income paid in kind (a) Main occupation Value of payment in fonn of food,cops, animals Value of paymnt in form of subsidized housing Value of payment in form of subsidized tranwport Value of payment In other forms (a) Employment income paid in kind, main occupation (b) Seconday occupation Value of payment in fonn of food, crops, animals Value of payment in other forms (b) Employment income paid in kind, secondary occupation (c) Other occupations Value of payment in kind: third occupation Value of payment in kind: fourth occupation Value of payment in kind: fifth occupation (c) Employment income paid in kind, other occupations a a a a a a a a a Expenditure corresonding to employment income paid in kind - (a) + (b) + (c) Definiti of the calculation categories a. Reate to the calculaton of most of the empoyment income categoies, where the wage rate (in csh or kind) is conved Into a weeldy rate and multiplied by the number of weeks the respondent worked in the past year. However, if the wage rate is expressed on an annual basis, no fuher trsformatio is requied. b. Reates to componens lculated using Inomation not included in the questionnaire In practice this relates to estimates of hau and employees social security payments made by dtose quotig their wages post-ax. c. Relates to the valuation of harvested crops used in payment, kept as smed, used for processing etc. The quantity in a specific referene period is multiplied by a unit value and by365 and divided by the number of days in the reerene period. The unit value is estimated for each household by dividing sales revenue by the quantity of the crop sold. d. Relates to the calculation of a number of revenues or expenitures in which payment in a spedfied referec peiod is multiplied by365 divided bythe number ofdays in the reference a. Relates to the valuation of the consumption of domestic output of food, in which the frequency of consumption in a specified reference period is multiplied by the typical value of consumption on each occasion and by 365 and divided by the number of days in the refrence period and by the number of monwts in which an item is consumed, expresd as a proportion of 12. f Relates to the estimation of the value of depeiation of producer durable goods baxed on their cuent value and the estimated depreciation rate. Specfic methods of undertaldng this ar descibed in the text. g. Reates to many of itums in the production account of non-farm entises and to the majory of ependitumes. The calculation Is of the same type as 'e above except that, additionally, an allowance for seasonsaity is made by multiplying by the number of months 236

241 In which the anterise opeated or in which a pwtcoar wunm preuedas A propordt of 12. item was pumcsed, h. Relaes totheestimaton of imputad et forowneroccupid dweiigbased on hedonic equations reatin the'annual vaue of rent In cash and kindt the charact_st and amdidtles of the dwelig. See text for futher detail Reates eesmation of use values or Imputed consumption fow of durable goods This may be es6mated as 'T above, althgh In plndple an lnterst opprtunity cst showuld belmduced. See tet for further detais *~~~~~~~~ J l~~~~~~~~~~~~~~~~~~s \ 1

242 Bibliography AddisoT Employment Policy ande Nutritional Status n Ghan and nomic Adjustment in The Gambia. Report its Detenninants. SDA Working Paper 3. for an ILO Preparatory Assistance Mission. Washington, D.C.: World Bank. ELO, Geneva (March). Alessie, RI, P. Baker, R. Blundell, C Heady, and Labor Markets and Employ- C Megfr The Working Behaviour of ment Under Adjustment." SDA Project YoungPeople in Rural C8tedIvoire: Evidence Analysis Plan. World Bank, Africa Region, from the LSMS Pani" University College, SDA Unit, Washngton, D.C. London. Addison, T., and L Pmery "Macro- Appleton, S., P. Coier, and PRH HorsL Econooic Stabilization, Income Distribution 'GenderDifferesinEducationalttanmt and Povertr- a Prelimnary Surveyn Over- and in labor Market Participation." World seas Development Institute, London. Bank, Africa Region, SDA Unit, Washington, Addison, T., L. Demery,J. Round and M. Ferroni. D.C Adjusting for the Poor: A Frameok for "Women and Devdeoptent Policy Actions in A,4'r. Washington, D.C.: CotedIvire." Oxford Unrversity, Oxford. World Bank (fortcoming). Arida, P., and L. Taylor "Short-Run Addison, T., L. Demeiy, J. Round, T. Stephens, Macroeconomics.a In H. Chenery and T. and M. Fend Strcturl Adjustment and Srinivasan, eds. Handbook of Deelpment Eco- Poerty: A Conctal, Empirical and Policy nomics,vol.il Amsterdam: North-Holiand. Fmework. Washington, D.C.: World Bank. Atkinson, A.B "On the Measurement of Adelman,., and S. Robinson "Macro- Povertly Ecnmetrica 55: econwmic Adjustment and Income Distribu- Azam, J-P., et al "The Impact of Macrotion: Alternative Models Applied to Two economic Pohcies oh the Rual Poor." UNDP Economies." Journal of Deveopmnent Economics Policy Discussion Paper. UNDP, New York. 29(1): Bardhan, P. K Land, Labor, and Rural Pov- Ahned, R., and N. Rustagi "Marketing erty. New York. Columbia University Press. and Price Incentives in African and Asian Barnum,Howa*dN.,andLynSquire "An Countries: A Comparison." In D. Elz, ed., Econometric Application of the Theory of AgrculturalMetng Strate and Prcng Pol- Farm-Household. Journ of Devlpmen Ecoicy. Washington D.C.: World Bank. nomiss 6(1) Alderman,'Harold "Cooperative lairy Barera, A "The Role of Matemal School- Development in Karnataka, India: An Assess- ing and its Interactions with Public Health ment.~' International Food Policy Research Progms in Child Health Production." JDE. Institute, Washington, D.C pp

243 Becker, C A Theory of the Alloation of Analytical Basisfor the UNICEPAdvo=a and Time." Ecnomic Jooul a Case Comparison." In John CaIdwell, ed., Behrman, Jere R The Impact of ECD- Cultunil, Sodal nd Bhavio Deteminants of nomic Adjustment PrOgrams on Health and Healtih What is the Evidence? Canberra, Aus- Nutrition in Developing Countries. In David tralia: Australian National University. E. Bell and Michael R. Riech, eds., Health, Nu- Behrman, Jere R., A. Deolalikar, and B. Wolfe. trition and Economi Cris: Approce to Policy Nutrlients. Impacts and Determinants. in the Third World. Dover, MA. Aubum House. World Bank Econmic Review a. "lntrahouseiold Alction of Behnan, JereR.,andChaongphobSussangkarn. Nutrients and Gender Effects." In Siddig R "ParentlSdhoolgandChildOutcomes: Osmani, ed., Nutrition ad Poterty. Oxford Mother Versus Father, Sdhooling Quality, and Oxford University Press. Interactions.- University of Pennsylvania,. 1988b. "Intrahousehold Allocation of Philadelphia. Nutrients in Rural India: Are Boys Favored? Behmnan, Jere R., an Barbara L. Wolfe. 1984a. Do Parents Exhibit Inequality Aversion?" Ox- "The Socioenmic hnpact of Schooling in a ford Economic Pees. Developing Country." ReviW of Fxomics and _ 'Interactions Among Human Staistics 66(2): Resources and Poverty: What We Know and. 1984b. "More Evidence on Nutrition What We Do Not Know." ISMS Working Demand: Income Seems Overrated and Paper Series. World Bank, Washington, DlC Women's Schooling Underemphasized." Behrman,Jere R, and Nancy Birdsall OTe ourna of Del:opment Ecwonocs 14. Quality of Schooling; the Standard Focus on Benjamin, Dwayne Household Compo- Quantity Alone is Misleading." Amerian Eco- sition and Labor Demand: A Test of Rural nomi Review 73(December): Labor Market Efficiency." Researh Pogram Behbman,JereR,and Anil Deolalikar "Will in Development Studies, Discussion Paper 140. Developing Country Nutrition Improve with Woodrow Wilson School, Princeton Univer- Income? A Case Study for Rural South India.0 sity, Princeton, NJ (November). Journa of Political Economy 95: Berthdemny, J.C, and F. Cagey The Agri a. "Unobserved Household and cultural Supply Prike Elasticity in Africa, A Community Heterogeneity and the Labor Note on Peasants' Rationality in Non-Walra- Market Impact of Schooling. A Case Study for sian Context Euopamn Ewmosics Revwe31(8). Indonesia.' University of Pennsylvania, Besley, T.J, and S.M.R. Kanbur "Food Philadelphia. Subsidies and Poverty Alleviaton. Eonomic. 1988b. "Impact of Macro Economic Jounal 98. Adjustment on the Poor and on Social Sectors Bevan, D. L, A. Bingst, and P. Coller in Jamaica.1 University of Pennsylvania, 'Peasant Supply Response in Radoned Econo- Philadelphia (prepared for World Bank, Op- mies." World Delopmet 45(4). erations Evaluation Department). Bevan, D., P. Collier, and J. Gunning; with A c. "Health and Nutrition. In Bigsten and P. Horsnell Peasants and Hollis B. Chenery and TN. Srinivasan, eds., Goemnts. Oxford Oxford UniversityPress. Handbook on Economic Development, Vol. 1. Binswanger, H How Agricultural Pro- Amsterdam: North Holland. ducemrs Respond to Prices and Government In a "School Repetition, Dropouts vestnents. Paper preted at the First Anand the Returns to Schooling: The Case of In- nual World Bank Conferenceon Deveopment donesia" Uiversity of Pennsylvania, Phila- Economics, April, Washington, DC. delphia. Birdsall, Nancy "Public Inputs and Chld. 1989b. 'Wages and Labour Supply Schooling in Brazil." Journl of Deveoent in Rural India: The Role of Health, Nutrition Economics 18(1).67-8 and Seasonality." In D. Sahn, ed., Cause and BirdsallNancy,andJereR.Behnan "Does Implications of Samsonal Variaiy in Household Geogaphical Agrgtion Cause Overesti- Food Security. Baltimore: Johns Hopkins Uni- mates of the Returns to Sdcholing?" OXfod versity Press. Buletin of Eqmomi ad Stattics 46: "Health, Nutrition and Macro- Blaug, M "An Economc Analysis of Pereconomic Ad justient with a Human Face: The soral Earnings in Thailand." Economi Dead- 240

244 opmn and Cultural Change (October1-a sia" Journa of Politil Economy 80. Blundel, R "Consumer Bebaviour. Te- Casley, D.J., and D.A. Lty Data Collection oty and Empiricl Evidence - A Surv. in Dloping Coun Oxford: Carendon Ecnomic Joun 98. Press. Blundell,, and L. Walker "Modelling the Chadeau, A., and C. Roy "Relating Joint Determination of Household Labour Househoda Fina Consumption to Household SuppliesandConMOdityDemands." Economic Activitle:SubstitutabltyorComplmentarity Journal 92: betweenmarketandnon-marketproducfion." Boateng, E.O., K. Ewusi, R. Kanbur, and A. Reiew of lncome and Walth, Senes 32: McKay "A Poverty Profile for han, Chambers, R "SustaIable Rural Liveli &8 World Bank, Africa Region, SDA hoods: A Key Strategy for People, Environ- Unit, Washington D.C (September). ment and Developmntm In C. CoWoy and Boissiere, M., J.B. Kight, and RH. Sabot M. Utvinoff, eds, 7The Greening of Aid. Sus- "Earndngs, Schooling, Ability and Cognitive tainable Livelihoods in Practice. London: Skills." American Economic Review 75: Earthscan, in assocdiaion with the Intermaiaonal Bond, M "Agricultural Respones to Prices Instiute for Environment and Development. in Sub-Saharan African Countries." IMF Staff Chambers, R., R. Longhur and A. Pacey, eds. Papeas Seona Dimsion to Rural Povety. Bouis, HE., and LJ. Haddad "Comparing Londonm Fanis Pinter. calorie-income elasticities using calories de- Cheney, H.B, M. S. Ahluwalla, CLG. Bell, J.H. rived from reported food purchases and a Duloy, and R. Jolly Reistribution with twenty-four hour recall of food intakes: An Growth. London: Oxford University Press. application using Philippine data." Develop Chiswick, CU "On simating Earnings ment Economics Research Centre Discussion Functions for LDCs.- Jounl of Devel nt Paper 88, University of Warwick, Wasrwic Economics lourguignon, F., W. Bnon, and J. de Melo "Analysis of Earnings from "Macro-Micro Simulation Model." Household Enterp : Methodology and OECD Development Centre, Paris. Application to Thailand." Reviw of Econmics Boyd, D The Impact of Adjustment Poli- and Statitics 65:658>62. cies on Vulnerable Groups: The Case of Ja- Collier, P Maco-eonmic Polic, Employmaica, In GA. Cornia, R. Jolly, men and Living Standards inmalaroi ad Tanzaand F. Stewart, eds., Adustment with a Hunmn ma, ILO World Employment Pro- Pacm Oxford: Clarendon Press. gramue Research, International Employment Braun, J. von 'Effects of Technological Policies, Working Paper No. 18 (WP2-46/ Chane in Agriculture on Food Conmption WP.18). and Nutrition: Rice in a West Africa Setting." "Women in Development De- World Delopnme 16. finig the Issues." WPS 129, World Bank, "CommercIalzation of Small- Washington, DC (December). holder Agriculture Policy Requirements for "Women under Structural Ad- Capturing Gains for the Maloried Poor." sment.e SDA Pjet Anaysis Plan. World Intenational Pood Policy Research Institute, Bank, Africa Region, SDA Unit, Washingon, Washngton, D.C D.C. Braun,J.von,D.Hotchkdss,andMaarte nk "Women and Structural Ad Nonaditiol Export Crops in Guatema justment. Wodd Bank, Economic Develop- Effects on Production, Income, and Nutrition. ment Institute, Washington, D.C (forthcom- Research Report No. 73. Washington, D.C-.. ig). Internaonal Food Policy Research Insttute. Collier, P. and D. LaL labor and Poverty in Braun, J. von, D. Puetz, and P. Webb Imri- Kemp, Oxford: Clarendon Press gation Technolgy and m ion f Rice Colier, P, 5. Radwan, and S Wangwe in the Gambia Effects of lncome and Nutrio Labour and Poey in Runl Tanzan. Oxford: Rearch Report No. 75. WashIngton, D.C: Oxford Universty Press. Internatal Food Policy Research Institute. Corrda, GCA, R Jolly, and P. Stewart, eds Bruno, M "Domestic Resource Costs and Adjustment with a Huawn Face: Vol. I- Pro- Effective Ptotection: Cwarification and Synthe- tcing the Vulnerable and Promoting Growth. 241

245 Oxford: Clarndon Press for UNICEF. Review Economics and Stas., Adjustment with a Human Face Dervis, K., J. de Melo, and S. Robnon Vol. 2- Country Case Studies. Oxford Clar- Generquilbitrum Models for Deveopment Polendon Press for UNICEF. icy. Cambridge: Cambridge University Press. Cornia, G.A., and F. Stewart 'Country Devereux, S., and R. Hay "Origin of Expeuience with Adjustment!' In GA. Comia, Fanine.- Manuscript, Food Studies Group, R Jolly, and F. Stewart (1987). Oxford. Dankelman,I.,andJ. Davidson Women and Edirisinghe, A he Food Samp Schem in Environent in the Third World: Aliancefor the Sn Lanka Costs, Benfits and Optons for Mdif- Future. London: Earthscan. cation. Research Report No. 58. Washington, Dasgupta, P., and D. Ray 'Inequality as a D.C: Intenational Food Policy Research In- Deteminant of Malnutrition and Unernploy- stitute. ment. Theory." Economic journal 96: Edwards,S ExchangeRateMisaignntein Deaton, A. 1987a. The Allocation of Goods within Developing Countris. World Bank Occasional the Houshold: Adults, Children, and Gender. Paper 2/New Series. Baltimore: Johns LivingStandardsMeasuementStudyWorking Hopkins University Press. Paper 39. Washington, D.C.: World Bank RealExchangeates,Devaluation. 1987b. "Estimadonofown-andcross- and Adjutent Exchange Rate Poliy h Dlevelprice elasticity from household survey data. oping Countries. Cambridge, MA: MlT Press. Journal of Econonmtrics 36. F.A.O Terms of Trade of Pakistan's Agricul "Quality, Quantity and Spatial lure. Varlationof Prices." Ameri Economic Review Ferni, M "The Urban Bias of Urban 78. Food Policy: Consequences and Alternatives.e Deaton, A., and D. Bennamin The Living Latin American Studies Program Dissertation Standards Surmey and Price Policy Ronm: A Stdy Series, Cornell University, Ithaca, NY. of Cocoa and Coee Production in COte d'ivoire. Ferroni, M., and R. Kanbur "Poverty Con- Uving Standards Measurement Study Work- scious Restructuring of Public Expenditure." ing Paper 44. Washington, D.C.: World Bank. Paper presented at the Africa Economic Issue Deaton, A., and A. Case Analysi of house- Conference, June4-7. Nairobi, Kenya, hold expenditures. Living Standards Measure- Folbre, Nancy "Hearts and Spades: Parament Study Worlkng Paper 28. Washington, digms of Household Economics.- World De- D.C.: World Bank. vekdme 14. Deaton, A., and J. Muellbauer Economics of Foster, JE "On Economic Poverty. A Sur- Consumer Bdeviour. Cambridge Cambridge University Press. vey of Aggregate Measures." Advances in Economerics 3: _ "On Measuring Child Costs; Poster, J.E., J. Greer, and E. Thorbecke "A with Applications to Poor Countries." Jounal Class of Decomposable Poverty Measurees of Political Economy 94(4): Econometrica 52: Demezy, L he Poverty Profile." SDA Foster, J.E., and A. Shorcks "Poverty Project Analysis Plan. World Bank, Africa Orderings." Ecometrica 56: Region, SDA Unit, Washington, D.C. Gertler, Paul, and Jacques van der Gaa Demety, L, and T. Addison. 1987a. 'Stabiliza- NTe Willingness to Pay for Medical Camr Evition Policy and Income Distribution in Devel- dence from Two Developing Countries. Baltioping Countries. World Development 15(12). more Johns Hopkins University Press Perty Alleviation Under Struc- Glewwe, P The Distribution of Welre in tural Adjushnent. Washington, D.C.: World COte d'ivoire in Living Standards Mea- Bank. surement Study Working Paper 29. Washing- Demery, D., and L. Demery ton, D.C. World Bank. "Macroeconomic Adjustment and Income Glewwe, P., and D. de Tray "The Poor Distribution in Malaysia, " OECD De- During Adjustment: A Case Study of C6te velopment Centre, Paris. d'ivoire." World Bank, Policy Plaming and Deolaikar, Ardl B "Do Health and Nutri- Research, Washington, D.C. don Influence Labor Productivity in Agricul- Glewwe, P., and J. van der Gaag 'Conture? Econometric EstiTmtes for Rural lndia." fronting Poverty in Developing Countries: 242

246 Definitions, IWonmation and Policies." World of Income and Wealdth, Series Bank, Washington, D.C. (December) Do it Yourself and GDP'. Greer,., and E. Thorbecke Food Poverty Reiew of Inome and Welth, Series 25: Profile Applied to Kenyan Smallholders." Horton, S Birth order and child nutri- Economic Development and Cultural Change tfonal status: Evidence frm the Philippines" 35(1): Economic Demlopment and Cultural Change 36. Griliches,Z Distributed Lags: A Survey." Horton, S., and B. Miller The Effect of Economdrica 35. Gender of Household Head on Expenditure: Grootaert, C "The Conceptual Basis of Evidence from Low Income Households in Ja- Measures of Household Welfare and their lm- waica.0 University of Toronto. plied Survey Data Requirements The Review Hsiao,C. 1986K Analysis of Panel Data. New York: of Income and Wealth, Series 29(1). Cambridge University Press he Role of Employment and lliffe, J he African Poor. Cambridge Earndngs in Analyzing Levels of iving. A Generl Canbridge University Press. Methodology with Applications to Maaysia and ILO Economally Acive Population, Esti- Thailand. Living Standards Measurement mates and Projections Geneva: Study Working Paper 27. Washington, D.C: International Labour Office. World Bank African Employmt Report Habidht, J.-P., et al "Height and weight Addis Ababa: jobs and Skills Programme for standards for pre-school children: Are there Africa. really ethnic differences in growth potential?" Jaeger, W., and C Humphreys The Effect Lat2:30. of Policy Reforms on Agicultural Incentves Hansen, S "Structural Adjustment Pro- in Sub-Saharan Africa" Amercan Journal of gram and Sustainable Development." Paper Agricultural Economic 70(5). prepared for the annual session of CIDIE, Au- Johnson, M., A.D. McKay, and J.L Round gust Washington, D.C. -Household Income and Expenditure Aggre- Harberge, A 'Economic Adjustment and gates from the Ghana Living Standards Surthe Real Exchange Rate." In S. Edwards and L. vey 1988: Methodology and Documentation." Ahamed, eds., Economic Adjustment and Ex- SDA Project in Ghana: Data Methodolog and change Rates in Developing Countries. Chicago: Processing Note Number 2, University of University of Chicago Press. Warwick. Harvey, C, ed Agricultural Priang Policy Johnson, O.E.G "Labor Markets, External in Africa. London: Macmillan. Developments, and Unemployment in Devel- Hausman, J., and W. Taylor 'Panel Data oping Countries." IMF Stff Studiesfor the World and Unobservable Effects." Econetrica 49. Economic Outlook (July): Hawrylyshyn, "Towards a Definition of Kakwani, N "Poverty and Economic Non-Market Activities." Review of Incmne and Growth with Applications to Cote d'ivoire. Wealth, Series 23: World Bank, Welfare and Human Resources Hay, Roge, and Graham Eele "Food Division, Washington, D.C Securityand Nutrition." SDA Prect Analysis Kanbur, S.M.R Measurement and Alle- Plan. World Bank, Africa Region, SDA Unit, viation of Poverty: with an Application to the Washington, D.C. Impact of Macroeconomic Adjustent.e 1MF Heller, P. S., A. L Bovenberg, T. Catsambas, K.- Staff Papers. Y. Chu, and P. Shome the Implicationof. 1988a. "Poverty and the Social Di- Fund-Supportd Adjustment Programs for Po- mensions of Structural Adjustment in CMte erty: Expeic in Selected Countries. IMF Oc- d'voire.- World Bank, Africa Region, SDA casional Paper 58. Washington D.C.: Intema- Unit, Washington, D.C. (Apil). tional Monetary Fund b. -rhe Implications of Adjust. Hicks, J.R The Social Framework. 4th ed. mentprograms forpoverty Conceptual Issues Oxford: Oxford University Press. and Analytical Framework. Paper presented Hicks, N., and A. Kubisch "Cutting Gov- to seminar on Implications of Adjstnent Proemment Expenditures in LDCs." Finance and grams for Poverty Groups, International Development (September). Monetary Fund, November. Washingtonl).C Hill, T.P "On Good and Services." Review _ Pov and the Soia Dinen- 243

247 sios of Strctumdl Mjustment in COte d'ivir and Applcations. Ansterdam: NOrth Hokand. SDA Working Paper 2. Washington, D.C.. Undauer, D. L, and R H. Sabot Te World Bank. Public/Private Wage Diffeietial in a Poor Katsell, L Devaluation: A Critical Ap- Urban Economy." Mounal of Deve&ipment Ewoproach of the IMF's Policy Presciption." noms 12(Fenary-April): Amerkan Economic Reviw 73(2): Lckwood, M., arnd P. Coler "Matenal Katz, A.J "Valuing the Sevices of Con- Education and the Vicous Cirle of High PersuwerDurables. Reviw oficme and Wallth, tility and Malnutrition: an Analytic Surveyr. Series World Bank Working Papers Series 130. Katz, Aj., and J. Pesidn "The Value of Washington, D.C. Services Provided by the Stock of Consumer Lucas,R On theteory of DRC Criteria." Durables, : An Opportunity Cost Mea- Jloal of Deueo nt Ecnmics 14. sure." Survey of Current Business 60(7): Maddala, GS limited-depeent and Kenedy, Eileen, and Bruce Cogill Income Qualitative Variables in Econometrics. Camand Nutitiona Effcts of the Cmmercalizition bridge: Cambridge University Press. of Agrculture in Soutwestern Kenya. Research Intr,dudion to Econometdcs. Report No.63. Washington,D.C: lntermational New York: Maman. Food Policy Research Institute. McDonald,.M., and R. M. Solow "Wages _ he Effects of Sugar Cane Pr- and Employment in a Sepgented Labor Marduction on Food Security, Halth, and Nution in ket. Quarterly Journal of Econois C(4):1115- Xen A LongtudWinal Anaysis. Research Re. 42. port No. 78. Washington D.C.: Interational Meerman,J PubRcExpditurin Maihayi Food Policy Research Institute. Who Befs and Why? New York. Oxford Klligsworth, M. R Labor Supply. Cam- University Press. bridge: Cambridge University Press. Mezzera, J Segmented Labour Markets Killingworth, M. R., and J. 1. Heckman Without Poicy-induced Labour Market Dis "emle Labor Supply: A Survey.- In 0. C tortons.v WorldDevekment 9(11/12): Ashenfelter and R. Layard, eds., Handbook of Neay, J., and K. Roberts The Theory of Labor Economics, Volume 1. Amsterdan: North Household Behavior Under Rationing. Euro Holland. pn Econmc Review 13(1). Knudsen, O., and P. Scandizzo '"Te De- Nerlove, M. 195 Estimtes of the Elasicities miand for Calories in Developing Countries.e of Supply of SeleceWd Agricultual Commodi- Anmen Joumnal ofagriculhal Eom s 64:80- ties. Journal of Fawm Economics Distributed Lags and Esti- Koyck, L Ditribte Lags and Inestmt tion of Long-run Supply and Demand Elastid- Anals. Amsterdaim: North Holland. ties: Theoretical Considerations." Journd of Krueger, A.O., Maurice Schiff, and Alberto Farm Econoics 40. Valdez "Agicultural Incentives in De- Ongao, W.A "Adoption of New Farming veloping Countries- Mesuring the Effect of Technolog a Case Study of Maize Produc- Sectoral and Economy-wide Policies.' World tion in Western K a. Ph.D. diss., Univer- Bank Economic Revie 2(3) sity of Gotebtu Kumnar, Shubh Impact of Subsidized Rce on Payne, P Food System Indicators." In A. Food Nutrition in Kba. Research Report No. Pacey and P. Payne, eds., Agricultural Dveop- 5. Washington, D.C. Internatdonal Food Poicy ment and Nutrition. London: Hutchison. Research Institute. Pearce, Ridcard "The Sodal Dimensions Kusnic, M.W, and 1. Da Vanzo Income In- of Adjustment in Angola." Food Studies euaity and Dgefton of Income The Cae of Group/UNICEF, Oxford. Malysa Santa Monica, CA: Rand. Pearce, Richard, and Joseph Mullen rhe Latham, M.C "Strategies for the Control Smalhider Secr Under Adjustment." SDA of Malnutrition and the Influence of the Nutri- Project Analysis Plan. World Bank, Africa tional Scces. Food and Nutrition 10(1):5-31. Region, SDA Unit, Wahington, D.C. Lau,L Applications of Prfit Functions Pinstrup-Andersen, P "An analytical In M. A. Fuss and D. L McFadden, eds., Pro- famework for diussing nutritional effects ducffon Ecnmics A Dud Aprah to Theoy ofpodcies and program" In CEC Mam and 24

248 B. Huddleston, eds., Food Policy. Mloomington: Bank, Policy, Planning and Research Working Indiana University Press. Paper on Agricultural Policies, WPS 167, "Government Policy, Food Se- Washngton, D.C (March). curity md Nutrition in Sub-Saharan Africa." Ravallion, M., and M. HuppL 'Poverty PEW/Cornell Lecture Series on Food and and Undemutrition in Indonesia During the Nutrition Policy, Itdaca, NY. 1980s!" Agriculture and Rural Development Pitt, Mark M "Fod Preferences and Nu- Departnent Working Paper 286. World Bak, tritlon in Rural Bangladesh." Review of Eco Washington, D.C. (October). nomics and Statistics 65: Ravallion,M,andD.vandeWalle Poverty "Access to and Use of Health Orderings of Food PricingRleorms. Universityof Services Under Adjustment. SDA Project Warwick, Development Economics Research Analysis Plan. World Bank, Africa Region, Centre Discussion Paper 86 (August). SDA Unit, Washington, D.C. Republic of Ghana Ghana Linig Standards Pitt, Marwk M., and Mark R. Rosenzweig Suroy: Preliminay Results, Statistical HealthandNutrientConsumptionAcrossand Service, Accra, and World Bank, SDA Unit, Within Farm Householdf Review of Economics Washington, D.C. and Statistics (May) Reutlinger, S., and A. Selowsky Malnutri- _ "Agricultural Prices, Food tion and Poverty. Magnitude and Policy Options. Consumption and the Health and Productiv- Baltimore: Johns Hopkins University Press. ity of Farmers." In Inderdjit Singh, L Squire, Richards, PJ Preserving Jobs Under Ecoand J. Strauss, eds., Agricultural Household nomic Stabilisation Programmws: Can There Models: Extensions and Applications. Baltimore: be an Employment Target?" Intrnational Labour Johns Hopkins University Press. Review 125(4) "The Selectivity of Fertility and Rosenzweig, Mark R., and T. Paul Schultz the Deteminants of Human Capital Invest- "Market Opportunities, Genetic Endownents, ments: Parametic and Semi-Parametric Esti- and Intrafamily Resource Distribution: Child mates. Brown University, Department of Survival in Rural India." Anmrican Econondc EconomicsWorking Paper Providence, Review RI (October) "Market Opportunities and "Estimating the Intrafamily Intrafamily Resource Distribution: Reply.' Incidence of Illness Child Health and Gender Amerian Economic Review 74(3): Inequality in the Allocation of Time in Indo- Ruggles, R., and N. D. Ruggles The nia Intmatinl Economic Review (forth- Integration of Macro and Micro Data for the coming). Household Sector." Review of Income and Pitt, Mark M., Mark R. Rosenzweig, and Md. Wealth, Series Nazmul Hassan "Productivity, Health Sahn, D. E "The Effect of Price and Income and Inequality in the Intrahousehold Distribu- Changes on Food-Energy Intake in Sri-Lanka." tion of Food in Low-Income Countries." Economic Devepment and Cultural Change. Amerkan Ecnomic Review (forthcoming) Malnutrition in COte d'ivoire: Psacharopoulos,G "Retums toeducation: Prevalen and Detemnns. SDA Working A Furher International Update and Implica- Paper 4. Washington, D-C: World Bank. tions. Journal of Human Resources Sahn, David, and Harold Alderman "The * "Education and Development Effect of Human Capital on Wages, and the A Review." World Bank Research Observer Deterninants of Labor Supply in a Develop- 3(1): ing Country." Journal of Dvelopmt Econom- Pudney, S Modeling Individual Choice Ox- ics. ford: Basil Blackwell. Saunders, CT "Measures of Total House- Pyatt,G AAccounting fortinte Use." Revkw hold Consumption. Review of Income and of Income and Wealth, Series 36: Weath, Series 26: Pyatt, G., and E. Thorbecke Planning Scandizzo, P., and C Bruce Methodologies.chniques for a Better Future. Geneva.: Inter- for Measunng AgriculturalPrice Interenton Efnadonal Labour Office. feds. World Bank Staff Working Paper 394. Ravallion,M DoPricelncreases for Staple Washington, D.C: World Bank. Foods Help or Hurt the Rural Poor?' World Schultz, T. Paul. 1988a. "Education Investments 245

249 and Returns." In Hollis Chenery and T.N. Strauss, John, and Duncan Thonas 'The Srinivasan, eds., Handbook of Devedopment Eco- Shapeof theexpendie-cloiecurve." Yale nomics. Amsterdam: North-Holland. University, New Haven, CT b. 'Women and Development Svedber, P "Undernutrtiton in Sub-Sa- Objectives, Framework, and Policy Intervenr haran Africa: Is There a Sex Bias?" Institute tions. Yale University, New Haven, Cr. for Intemational Economic Studies Seminar Scobie, G.M "Macroeconomic Adjustment Paper 421, Stockholm. and the Poor: Toward a Research Strategy." Thomas, V., J. Strauss, and M. HendrAques CFNPP Macrograph 89-t, Ithaca, NY. "Child Survival, Height for Age and House- Scott, C Testing Survey Methodology in hold Characteristics in Brazil." ounal of Dethe Context of the SDA Program." World Bank, vedopment Econmmics. Africa Region, SDA Unit, Washington, D.C. Thorbecke, E 'Adjustnent, Growth and Estimation of Annual Expen- Icome Distribution in Indonesia.' OECD Dediture From One-Month Cross- Sectional Data velopment Centre, Paris. in a Household Survey." World Bank, Africa Timnmer, C. P "Is There'Curvatre' in the Region, SDA Unit, Washington, D.C. (June). Slutsky Matrix? Review of Economics and Sta- Sen, A Poverty and fiines: An essay on tistics 62: exchangeenttlements. Oxford: Clarendon Press. Timmer, C. P., and H. Alderman "Esti "Poor Relatively Speaking. mating Consumption Parameters for Food Oxford Economic Papers 35IJuly): Policy Analysis.' American Journd of Agricul he Standard of Living. Cambr- urui Economtics 61(5) idge: Cambidge University Press. Timmer, CP., W.P. Falcon, and SR. Peson Sen. A., and S. Sengupta "Malnutrition of Food Policy Anaysis. Baltimore Johns Hopkins Rural Children and the Sex Bias." Economic and University Press. Politica Weky 18: Tfinenburg K "Experience with Food Singh, I., L. Squire, and J. Strauss. 1986a. Agri- Strategies in Four African Counties. In J.P. culturl Household Models: Extesions, Applica- Gittinger, J. Leslie, and C Holsington, eds., lions and Policy. Baltimore: Johns Hopkins Food Policy: Interating Supply, Distribution and University Press. Consumption. Baltimore: Johns Hopkins UTd "A Survey of Agricultural versity Press. Household Models: Recent Fmdings and Pol- Turvey, R., ed Deelpments in Inteaicy Implications." World Bank Econom Review tinl Labour Statistics. London: Pinter Publ(1): lishers for ILO. Strassmann, W.P "ome-based Enter- United Nations, National Household Survey Caprises in Cities of Developing Countries." pability Prrame Household Income Economc Deelopnent and Cultuml Change 36(1) and Expenditure Surveys. (Sri Lanka). United Nations Statistical Office A System Strauss,John "Joint Determination of Food of Wationd Accounts. Series F, No. 2, Rev. 3. Consumption and Production in Rural Sierra New York. United Nations. Leone: Estimates of a Household Firm Model.V van der Hoeven, R "External Shocks and Journal of Development Economics 14: Stabilisation Policies: Spreading the Load." _ "Does Better Nutrition Raise Intermaional labour Reviw 126(2): Farm Productivity?" Journal of Political Verma, V., T. Marchant, and C Scott Econmy 94. "Evaluation of crop-cut methods and farmer. 1988a. 7he Effects of Household and reports for estimating crop production: Re- Community Characris on the Nutrition of sults of a methodological study in five African Preschool hildren: Evidence from RuraI C8te countries.' Longacre Agricultural Developd'Ivoire. LivingStandardsMeasurenentStudy ment Centre Ltd., London. Working Paper 40. Washingtn, D.C.: World Vijverberg, W. 1988a. Profits fim Self-Employ- Bank. ment. A Case Study of COte d'ivoire. Living "Households, Communities, StandardsMeasuremnentStudyWorkingPaper and the Nutrition of Pr-School Children: Evi- 43. Washington D.C.: World Bank. dence from Rural Cote d'lvoire." Economic. 1988b. NonagricWulnlFamly Enter- Devlpment and Cultural Change 38. prises in COle d'lwirc A Descriptive Analysi. 246